]>
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 | |
1f55a6ec AV |
495 | struct nameidata { |
496 | struct path path; | |
caa85634 AV |
497 | union { |
498 | struct qstr last; | |
499 | struct path link; | |
500 | }; | |
1f55a6ec AV |
501 | struct path root; |
502 | struct inode *inode; /* path.dentry.d_inode */ | |
503 | unsigned int flags; | |
504 | unsigned seq, m_seq; | |
505 | int last_type; | |
506 | unsigned depth; | |
5e53084d | 507 | struct file *base; |
697fc6ca AV |
508 | struct saved { |
509 | struct path link; | |
510 | void *cookie; | |
511 | const char *name; | |
512 | } stack[MAX_NESTED_LINKS + 1]; | |
1f55a6ec AV |
513 | }; |
514 | ||
19660af7 | 515 | /* |
31e6b01f | 516 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
517 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
518 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
519 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
520 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
521 | * got stuck, so ref-walk may continue from there. If this is not successful | |
522 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
523 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 524 | */ |
31e6b01f NP |
525 | |
526 | /** | |
19660af7 AV |
527 | * unlazy_walk - try to switch to ref-walk mode. |
528 | * @nd: nameidata pathwalk data | |
529 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 530 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 531 | * |
19660af7 AV |
532 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
533 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
534 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 535 | */ |
19660af7 | 536 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
537 | { |
538 | struct fs_struct *fs = current->fs; | |
539 | struct dentry *parent = nd->path.dentry; | |
540 | ||
541 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 LT |
542 | |
543 | /* | |
48a066e7 AV |
544 | * After legitimizing the bastards, terminate_walk() |
545 | * will do the right thing for non-RCU mode, and all our | |
546 | * subsequent exit cases should rcu_read_unlock() | |
547 | * before returning. Do vfsmount first; if dentry | |
548 | * can't be legitimized, just set nd->path.dentry to NULL | |
549 | * and rely on dput(NULL) being a no-op. | |
e5c832d5 | 550 | */ |
48a066e7 | 551 | if (!legitimize_mnt(nd->path.mnt, nd->m_seq)) |
e5c832d5 | 552 | return -ECHILD; |
e5c832d5 | 553 | nd->flags &= ~LOOKUP_RCU; |
15570086 | 554 | |
48a066e7 AV |
555 | if (!lockref_get_not_dead(&parent->d_lockref)) { |
556 | nd->path.dentry = NULL; | |
d870b4a1 | 557 | goto out; |
48a066e7 AV |
558 | } |
559 | ||
15570086 LT |
560 | /* |
561 | * For a negative lookup, the lookup sequence point is the parents | |
562 | * sequence point, and it only needs to revalidate the parent dentry. | |
563 | * | |
564 | * For a positive lookup, we need to move both the parent and the | |
565 | * dentry from the RCU domain to be properly refcounted. And the | |
566 | * sequence number in the dentry validates *both* dentry counters, | |
567 | * since we checked the sequence number of the parent after we got | |
568 | * the child sequence number. So we know the parent must still | |
569 | * be valid if the child sequence number is still valid. | |
570 | */ | |
19660af7 | 571 | if (!dentry) { |
e5c832d5 LT |
572 | if (read_seqcount_retry(&parent->d_seq, nd->seq)) |
573 | goto out; | |
19660af7 AV |
574 | BUG_ON(nd->inode != parent->d_inode); |
575 | } else { | |
e5c832d5 LT |
576 | if (!lockref_get_not_dead(&dentry->d_lockref)) |
577 | goto out; | |
578 | if (read_seqcount_retry(&dentry->d_seq, nd->seq)) | |
579 | goto drop_dentry; | |
19660af7 | 580 | } |
e5c832d5 LT |
581 | |
582 | /* | |
583 | * Sequence counts matched. Now make sure that the root is | |
584 | * still valid and get it if required. | |
585 | */ | |
586 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
587 | spin_lock(&fs->lock); | |
588 | if (nd->root.mnt != fs->root.mnt || nd->root.dentry != fs->root.dentry) | |
589 | goto unlock_and_drop_dentry; | |
31e6b01f NP |
590 | path_get(&nd->root); |
591 | spin_unlock(&fs->lock); | |
592 | } | |
31e6b01f | 593 | |
8b61e74f | 594 | rcu_read_unlock(); |
31e6b01f | 595 | return 0; |
19660af7 | 596 | |
e5c832d5 LT |
597 | unlock_and_drop_dentry: |
598 | spin_unlock(&fs->lock); | |
599 | drop_dentry: | |
8b61e74f | 600 | rcu_read_unlock(); |
15570086 | 601 | dput(dentry); |
d0d27277 | 602 | goto drop_root_mnt; |
e5c832d5 | 603 | out: |
8b61e74f | 604 | rcu_read_unlock(); |
d0d27277 LT |
605 | drop_root_mnt: |
606 | if (!(nd->flags & LOOKUP_ROOT)) | |
607 | nd->root.mnt = NULL; | |
31e6b01f NP |
608 | return -ECHILD; |
609 | } | |
610 | ||
4ce16ef3 | 611 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 612 | { |
4ce16ef3 | 613 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
614 | } |
615 | ||
9f1fafee AV |
616 | /** |
617 | * complete_walk - successful completion of path walk | |
618 | * @nd: pointer nameidata | |
39159de2 | 619 | * |
9f1fafee AV |
620 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
621 | * Revalidate the final result, unless we'd already done that during | |
622 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
623 | * success, -error on failure. In case of failure caller does not | |
624 | * need to drop nd->path. | |
39159de2 | 625 | */ |
9f1fafee | 626 | static int complete_walk(struct nameidata *nd) |
39159de2 | 627 | { |
16c2cd71 | 628 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 629 | int status; |
39159de2 | 630 | |
9f1fafee AV |
631 | if (nd->flags & LOOKUP_RCU) { |
632 | nd->flags &= ~LOOKUP_RCU; | |
633 | if (!(nd->flags & LOOKUP_ROOT)) | |
634 | nd->root.mnt = NULL; | |
15570086 | 635 | |
48a066e7 | 636 | if (!legitimize_mnt(nd->path.mnt, nd->m_seq)) { |
8b61e74f | 637 | rcu_read_unlock(); |
48a066e7 AV |
638 | return -ECHILD; |
639 | } | |
e5c832d5 | 640 | if (unlikely(!lockref_get_not_dead(&dentry->d_lockref))) { |
8b61e74f | 641 | rcu_read_unlock(); |
48a066e7 | 642 | mntput(nd->path.mnt); |
e5c832d5 LT |
643 | return -ECHILD; |
644 | } | |
645 | if (read_seqcount_retry(&dentry->d_seq, nd->seq)) { | |
8b61e74f | 646 | rcu_read_unlock(); |
e5c832d5 | 647 | dput(dentry); |
48a066e7 | 648 | mntput(nd->path.mnt); |
9f1fafee AV |
649 | return -ECHILD; |
650 | } | |
8b61e74f | 651 | rcu_read_unlock(); |
9f1fafee AV |
652 | } |
653 | ||
16c2cd71 AV |
654 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
655 | return 0; | |
656 | ||
ecf3d1f1 | 657 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
658 | return 0; |
659 | ||
ecf3d1f1 | 660 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
661 | if (status > 0) |
662 | return 0; | |
663 | ||
16c2cd71 | 664 | if (!status) |
39159de2 | 665 | status = -ESTALE; |
16c2cd71 | 666 | |
9f1fafee | 667 | path_put(&nd->path); |
39159de2 JL |
668 | return status; |
669 | } | |
670 | ||
2a737871 AV |
671 | static __always_inline void set_root(struct nameidata *nd) |
672 | { | |
7bd88377 | 673 | get_fs_root(current->fs, &nd->root); |
2a737871 AV |
674 | } |
675 | ||
7bd88377 | 676 | static __always_inline unsigned set_root_rcu(struct nameidata *nd) |
31e6b01f | 677 | { |
7bd88377 AV |
678 | struct fs_struct *fs = current->fs; |
679 | unsigned seq, res; | |
c28cc364 | 680 | |
7bd88377 AV |
681 | do { |
682 | seq = read_seqcount_begin(&fs->seq); | |
683 | nd->root = fs->root; | |
684 | res = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
685 | } while (read_seqcount_retry(&fs->seq, seq)); | |
686 | return res; | |
31e6b01f NP |
687 | } |
688 | ||
1d957f9b | 689 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
690 | { |
691 | dput(path->dentry); | |
4ac91378 | 692 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
693 | mntput(path->mnt); |
694 | } | |
695 | ||
7b9337aa NP |
696 | static inline void path_to_nameidata(const struct path *path, |
697 | struct nameidata *nd) | |
051d3812 | 698 | { |
31e6b01f NP |
699 | if (!(nd->flags & LOOKUP_RCU)) { |
700 | dput(nd->path.dentry); | |
701 | if (nd->path.mnt != path->mnt) | |
702 | mntput(nd->path.mnt); | |
9a229683 | 703 | } |
31e6b01f | 704 | nd->path.mnt = path->mnt; |
4ac91378 | 705 | nd->path.dentry = path->dentry; |
051d3812 IK |
706 | } |
707 | ||
b5fb63c1 CH |
708 | /* |
709 | * Helper to directly jump to a known parsed path from ->follow_link, | |
710 | * caller must have taken a reference to path beforehand. | |
711 | */ | |
712 | void nd_jump_link(struct nameidata *nd, struct path *path) | |
713 | { | |
714 | path_put(&nd->path); | |
715 | ||
716 | nd->path = *path; | |
717 | nd->inode = nd->path.dentry->d_inode; | |
718 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
719 | } |
720 | ||
b9ff4429 | 721 | static inline void put_link(struct nameidata *nd) |
574197e0 | 722 | { |
b9ff4429 AV |
723 | struct saved *last = nd->stack + nd->depth; |
724 | struct inode *inode = last->link.dentry->d_inode; | |
725 | if (last->cookie && inode->i_op->put_link) | |
726 | inode->i_op->put_link(last->link.dentry, last->cookie); | |
727 | path_put(&last->link); | |
574197e0 AV |
728 | } |
729 | ||
561ec64a LT |
730 | int sysctl_protected_symlinks __read_mostly = 0; |
731 | int sysctl_protected_hardlinks __read_mostly = 0; | |
800179c9 KC |
732 | |
733 | /** | |
734 | * may_follow_link - Check symlink following for unsafe situations | |
735 | * @link: The path of the symlink | |
55852635 | 736 | * @nd: nameidata pathwalk data |
800179c9 KC |
737 | * |
738 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
739 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
740 | * in a sticky world-writable directory. This is to protect privileged | |
741 | * processes from failing races against path names that may change out | |
742 | * from under them by way of other users creating malicious symlinks. | |
743 | * It will permit symlinks to be followed only when outside a sticky | |
744 | * world-writable directory, or when the uid of the symlink and follower | |
745 | * match, or when the directory owner matches the symlink's owner. | |
746 | * | |
747 | * Returns 0 if following the symlink is allowed, -ve on error. | |
748 | */ | |
749 | static inline int may_follow_link(struct path *link, struct nameidata *nd) | |
750 | { | |
751 | const struct inode *inode; | |
752 | const struct inode *parent; | |
753 | ||
754 | if (!sysctl_protected_symlinks) | |
755 | return 0; | |
756 | ||
757 | /* Allowed if owner and follower match. */ | |
758 | inode = link->dentry->d_inode; | |
81abe27b | 759 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
760 | return 0; |
761 | ||
762 | /* Allowed if parent directory not sticky and world-writable. */ | |
763 | parent = nd->path.dentry->d_inode; | |
764 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) | |
765 | return 0; | |
766 | ||
767 | /* Allowed if parent directory and link owner match. */ | |
81abe27b | 768 | if (uid_eq(parent->i_uid, inode->i_uid)) |
800179c9 KC |
769 | return 0; |
770 | ||
ffd8d101 | 771 | audit_log_link_denied("follow_link", link); |
800179c9 KC |
772 | path_put_conditional(link, nd); |
773 | path_put(&nd->path); | |
774 | return -EACCES; | |
775 | } | |
776 | ||
777 | /** | |
778 | * safe_hardlink_source - Check for safe hardlink conditions | |
779 | * @inode: the source inode to hardlink from | |
780 | * | |
781 | * Return false if at least one of the following conditions: | |
782 | * - inode is not a regular file | |
783 | * - inode is setuid | |
784 | * - inode is setgid and group-exec | |
785 | * - access failure for read and write | |
786 | * | |
787 | * Otherwise returns true. | |
788 | */ | |
789 | static bool safe_hardlink_source(struct inode *inode) | |
790 | { | |
791 | umode_t mode = inode->i_mode; | |
792 | ||
793 | /* Special files should not get pinned to the filesystem. */ | |
794 | if (!S_ISREG(mode)) | |
795 | return false; | |
796 | ||
797 | /* Setuid files should not get pinned to the filesystem. */ | |
798 | if (mode & S_ISUID) | |
799 | return false; | |
800 | ||
801 | /* Executable setgid files should not get pinned to the filesystem. */ | |
802 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
803 | return false; | |
804 | ||
805 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
806 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
807 | return false; | |
808 | ||
809 | return true; | |
810 | } | |
811 | ||
812 | /** | |
813 | * may_linkat - Check permissions for creating a hardlink | |
814 | * @link: the source to hardlink from | |
815 | * | |
816 | * Block hardlink when all of: | |
817 | * - sysctl_protected_hardlinks enabled | |
818 | * - fsuid does not match inode | |
819 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
820 | * - not CAP_FOWNER | |
821 | * | |
822 | * Returns 0 if successful, -ve on error. | |
823 | */ | |
824 | static int may_linkat(struct path *link) | |
825 | { | |
826 | const struct cred *cred; | |
827 | struct inode *inode; | |
828 | ||
829 | if (!sysctl_protected_hardlinks) | |
830 | return 0; | |
831 | ||
832 | cred = current_cred(); | |
833 | inode = link->dentry->d_inode; | |
834 | ||
835 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
836 | * otherwise, it must be a safe source. | |
837 | */ | |
81abe27b | 838 | if (uid_eq(cred->fsuid, inode->i_uid) || safe_hardlink_source(inode) || |
800179c9 KC |
839 | capable(CAP_FOWNER)) |
840 | return 0; | |
841 | ||
a51d9eaa | 842 | audit_log_link_denied("linkat", link); |
800179c9 KC |
843 | return -EPERM; |
844 | } | |
845 | ||
3b2e7f75 AV |
846 | static __always_inline |
847 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 848 | { |
3b2e7f75 AV |
849 | struct saved *last = nd->stack + nd->depth; |
850 | struct dentry *dentry = nd->link.dentry; | |
0a959df5 | 851 | struct inode *inode = dentry->d_inode; |
6d7b5aae | 852 | int error; |
0a959df5 | 853 | const char *res; |
1da177e4 | 854 | |
844a3917 AV |
855 | BUG_ON(nd->flags & LOOKUP_RCU); |
856 | ||
3b2e7f75 AV |
857 | if (nd->link.mnt == nd->path.mnt) |
858 | mntget(nd->link.mnt); | |
0e794589 | 859 | |
3b2e7f75 AV |
860 | if (unlikely(current->total_link_count >= 40)) { |
861 | path_put(&nd->path); | |
862 | path_put(&nd->link); | |
863 | return ERR_PTR(-ELOOP); | |
864 | } | |
865 | ||
866 | last->link = nd->link; | |
867 | last->cookie = NULL; | |
6d7b5aae | 868 | |
574197e0 AV |
869 | cond_resched(); |
870 | current->total_link_count++; | |
871 | ||
3b2e7f75 | 872 | touch_atime(&last->link); |
cd4e91d3 | 873 | |
37882db0 | 874 | error = security_inode_follow_link(dentry); |
0a959df5 | 875 | res = ERR_PTR(error); |
6d7b5aae | 876 | if (error) |
0a959df5 | 877 | goto out; |
36f3b4f6 | 878 | |
86acdca1 | 879 | nd->last_type = LAST_BIND; |
d4dee48b AV |
880 | res = inode->i_link; |
881 | if (!res) { | |
3b2e7f75 | 882 | res = inode->i_op->follow_link(dentry, &last->cookie, nd); |
d4dee48b | 883 | if (IS_ERR(res)) { |
0a959df5 | 884 | out: |
d4dee48b | 885 | path_put(&nd->path); |
3b2e7f75 | 886 | path_put(&last->link); |
d4dee48b | 887 | } |
1da177e4 | 888 | } |
0a959df5 AV |
889 | return res; |
890 | } | |
6d7b5aae | 891 | |
31e6b01f NP |
892 | static int follow_up_rcu(struct path *path) |
893 | { | |
0714a533 AV |
894 | struct mount *mnt = real_mount(path->mnt); |
895 | struct mount *parent; | |
31e6b01f NP |
896 | struct dentry *mountpoint; |
897 | ||
0714a533 AV |
898 | parent = mnt->mnt_parent; |
899 | if (&parent->mnt == path->mnt) | |
31e6b01f | 900 | return 0; |
a73324da | 901 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 902 | path->dentry = mountpoint; |
0714a533 | 903 | path->mnt = &parent->mnt; |
31e6b01f NP |
904 | return 1; |
905 | } | |
906 | ||
f015f126 DH |
907 | /* |
908 | * follow_up - Find the mountpoint of path's vfsmount | |
909 | * | |
910 | * Given a path, find the mountpoint of its source file system. | |
911 | * Replace @path with the path of the mountpoint in the parent mount. | |
912 | * Up is towards /. | |
913 | * | |
914 | * Return 1 if we went up a level and 0 if we were already at the | |
915 | * root. | |
916 | */ | |
bab77ebf | 917 | int follow_up(struct path *path) |
1da177e4 | 918 | { |
0714a533 AV |
919 | struct mount *mnt = real_mount(path->mnt); |
920 | struct mount *parent; | |
1da177e4 | 921 | struct dentry *mountpoint; |
99b7db7b | 922 | |
48a066e7 | 923 | read_seqlock_excl(&mount_lock); |
0714a533 | 924 | parent = mnt->mnt_parent; |
3c0a6163 | 925 | if (parent == mnt) { |
48a066e7 | 926 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
927 | return 0; |
928 | } | |
0714a533 | 929 | mntget(&parent->mnt); |
a73324da | 930 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 931 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
932 | dput(path->dentry); |
933 | path->dentry = mountpoint; | |
934 | mntput(path->mnt); | |
0714a533 | 935 | path->mnt = &parent->mnt; |
1da177e4 LT |
936 | return 1; |
937 | } | |
4d359507 | 938 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 939 | |
b5c84bf6 | 940 | /* |
9875cf80 DH |
941 | * Perform an automount |
942 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
943 | * were called with. | |
1da177e4 | 944 | */ |
9875cf80 DH |
945 | static int follow_automount(struct path *path, unsigned flags, |
946 | bool *need_mntput) | |
31e6b01f | 947 | { |
9875cf80 | 948 | struct vfsmount *mnt; |
ea5b778a | 949 | int err; |
9875cf80 DH |
950 | |
951 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
952 | return -EREMOTE; | |
953 | ||
0ec26fd0 MS |
954 | /* We don't want to mount if someone's just doing a stat - |
955 | * unless they're stat'ing a directory and appended a '/' to | |
956 | * the name. | |
957 | * | |
958 | * We do, however, want to mount if someone wants to open or | |
959 | * create a file of any type under the mountpoint, wants to | |
960 | * traverse through the mountpoint or wants to open the | |
961 | * mounted directory. Also, autofs may mark negative dentries | |
962 | * as being automount points. These will need the attentions | |
963 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 964 | */ |
0ec26fd0 | 965 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 966 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
967 | path->dentry->d_inode) |
968 | return -EISDIR; | |
969 | ||
9875cf80 DH |
970 | current->total_link_count++; |
971 | if (current->total_link_count >= 40) | |
972 | return -ELOOP; | |
973 | ||
974 | mnt = path->dentry->d_op->d_automount(path); | |
975 | if (IS_ERR(mnt)) { | |
976 | /* | |
977 | * The filesystem is allowed to return -EISDIR here to indicate | |
978 | * it doesn't want to automount. For instance, autofs would do | |
979 | * this so that its userspace daemon can mount on this dentry. | |
980 | * | |
981 | * However, we can only permit this if it's a terminal point in | |
982 | * the path being looked up; if it wasn't then the remainder of | |
983 | * the path is inaccessible and we should say so. | |
984 | */ | |
49084c3b | 985 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
986 | return -EREMOTE; |
987 | return PTR_ERR(mnt); | |
31e6b01f | 988 | } |
ea5b778a | 989 | |
9875cf80 DH |
990 | if (!mnt) /* mount collision */ |
991 | return 0; | |
31e6b01f | 992 | |
8aef1884 AV |
993 | if (!*need_mntput) { |
994 | /* lock_mount() may release path->mnt on error */ | |
995 | mntget(path->mnt); | |
996 | *need_mntput = true; | |
997 | } | |
19a167af | 998 | err = finish_automount(mnt, path); |
9875cf80 | 999 | |
ea5b778a DH |
1000 | switch (err) { |
1001 | case -EBUSY: | |
1002 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1003 | return 0; |
ea5b778a | 1004 | case 0: |
8aef1884 | 1005 | path_put(path); |
ea5b778a DH |
1006 | path->mnt = mnt; |
1007 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1008 | return 0; |
19a167af AV |
1009 | default: |
1010 | return err; | |
ea5b778a | 1011 | } |
19a167af | 1012 | |
463ffb2e AV |
1013 | } |
1014 | ||
9875cf80 DH |
1015 | /* |
1016 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1017 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1018 | * - Flagged as mountpoint |
1019 | * - Flagged as automount point | |
1020 | * | |
1021 | * This may only be called in refwalk mode. | |
1022 | * | |
1023 | * Serialization is taken care of in namespace.c | |
1024 | */ | |
1025 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 1026 | { |
8aef1884 | 1027 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1028 | unsigned managed; |
1029 | bool need_mntput = false; | |
8aef1884 | 1030 | int ret = 0; |
9875cf80 DH |
1031 | |
1032 | /* Given that we're not holding a lock here, we retain the value in a | |
1033 | * local variable for each dentry as we look at it so that we don't see | |
1034 | * the components of that value change under us */ | |
1035 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1036 | managed &= DCACHE_MANAGED_DENTRY, | |
1037 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1038 | /* Allow the filesystem to manage the transit without i_mutex |
1039 | * being held. */ | |
1040 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1041 | BUG_ON(!path->dentry->d_op); | |
1042 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 1043 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 1044 | if (ret < 0) |
8aef1884 | 1045 | break; |
cc53ce53 DH |
1046 | } |
1047 | ||
9875cf80 DH |
1048 | /* Transit to a mounted filesystem. */ |
1049 | if (managed & DCACHE_MOUNTED) { | |
1050 | struct vfsmount *mounted = lookup_mnt(path); | |
1051 | if (mounted) { | |
1052 | dput(path->dentry); | |
1053 | if (need_mntput) | |
1054 | mntput(path->mnt); | |
1055 | path->mnt = mounted; | |
1056 | path->dentry = dget(mounted->mnt_root); | |
1057 | need_mntput = true; | |
1058 | continue; | |
1059 | } | |
1060 | ||
1061 | /* Something is mounted on this dentry in another | |
1062 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1063 | * namespace got unmounted before lookup_mnt() could |
1064 | * get it */ | |
9875cf80 DH |
1065 | } |
1066 | ||
1067 | /* Handle an automount point */ | |
1068 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
1069 | ret = follow_automount(path, flags, &need_mntput); | |
1070 | if (ret < 0) | |
8aef1884 | 1071 | break; |
9875cf80 DH |
1072 | continue; |
1073 | } | |
1074 | ||
1075 | /* We didn't change the current path point */ | |
1076 | break; | |
1da177e4 | 1077 | } |
8aef1884 AV |
1078 | |
1079 | if (need_mntput && path->mnt == mnt) | |
1080 | mntput(path->mnt); | |
1081 | if (ret == -EISDIR) | |
1082 | ret = 0; | |
a3fbbde7 | 1083 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
1084 | } |
1085 | ||
cc53ce53 | 1086 | int follow_down_one(struct path *path) |
1da177e4 LT |
1087 | { |
1088 | struct vfsmount *mounted; | |
1089 | ||
1c755af4 | 1090 | mounted = lookup_mnt(path); |
1da177e4 | 1091 | if (mounted) { |
9393bd07 AV |
1092 | dput(path->dentry); |
1093 | mntput(path->mnt); | |
1094 | path->mnt = mounted; | |
1095 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1096 | return 1; |
1097 | } | |
1098 | return 0; | |
1099 | } | |
4d359507 | 1100 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1101 | |
b8faf035 | 1102 | static inline int managed_dentry_rcu(struct dentry *dentry) |
62a7375e | 1103 | { |
b8faf035 N |
1104 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1105 | dentry->d_op->d_manage(dentry, true) : 0; | |
62a7375e IK |
1106 | } |
1107 | ||
9875cf80 | 1108 | /* |
287548e4 AV |
1109 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1110 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1111 | */ |
1112 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 1113 | struct inode **inode) |
9875cf80 | 1114 | { |
62a7375e | 1115 | for (;;) { |
c7105365 | 1116 | struct mount *mounted; |
62a7375e IK |
1117 | /* |
1118 | * Don't forget we might have a non-mountpoint managed dentry | |
1119 | * that wants to block transit. | |
1120 | */ | |
b8faf035 N |
1121 | switch (managed_dentry_rcu(path->dentry)) { |
1122 | case -ECHILD: | |
1123 | default: | |
ab90911f | 1124 | return false; |
b8faf035 N |
1125 | case -EISDIR: |
1126 | return true; | |
1127 | case 0: | |
1128 | break; | |
1129 | } | |
62a7375e IK |
1130 | |
1131 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1132 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1133 | |
474279dc | 1134 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1135 | if (!mounted) |
1136 | break; | |
c7105365 AV |
1137 | path->mnt = &mounted->mnt; |
1138 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1139 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 1140 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1141 | /* |
1142 | * Update the inode too. We don't need to re-check the | |
1143 | * dentry sequence number here after this d_inode read, | |
1144 | * because a mount-point is always pinned. | |
1145 | */ | |
1146 | *inode = path->dentry->d_inode; | |
9875cf80 | 1147 | } |
f5be3e29 | 1148 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1149 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1150 | } |
1151 | ||
31e6b01f NP |
1152 | static int follow_dotdot_rcu(struct nameidata *nd) |
1153 | { | |
4023bfc9 | 1154 | struct inode *inode = nd->inode; |
7bd88377 AV |
1155 | if (!nd->root.mnt) |
1156 | set_root_rcu(nd); | |
31e6b01f | 1157 | |
9875cf80 | 1158 | while (1) { |
31e6b01f NP |
1159 | if (nd->path.dentry == nd->root.dentry && |
1160 | nd->path.mnt == nd->root.mnt) { | |
1161 | break; | |
1162 | } | |
1163 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1164 | struct dentry *old = nd->path.dentry; | |
1165 | struct dentry *parent = old->d_parent; | |
1166 | unsigned seq; | |
1167 | ||
4023bfc9 | 1168 | inode = parent->d_inode; |
31e6b01f NP |
1169 | seq = read_seqcount_begin(&parent->d_seq); |
1170 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 1171 | goto failed; |
31e6b01f NP |
1172 | nd->path.dentry = parent; |
1173 | nd->seq = seq; | |
1174 | break; | |
1175 | } | |
1176 | if (!follow_up_rcu(&nd->path)) | |
1177 | break; | |
4023bfc9 | 1178 | inode = nd->path.dentry->d_inode; |
31e6b01f | 1179 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
31e6b01f | 1180 | } |
b37199e6 AV |
1181 | while (d_mountpoint(nd->path.dentry)) { |
1182 | struct mount *mounted; | |
1183 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
1184 | if (!mounted) | |
1185 | break; | |
1186 | nd->path.mnt = &mounted->mnt; | |
1187 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1188 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1189 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
f5be3e29 | 1190 | if (read_seqretry(&mount_lock, nd->m_seq)) |
b37199e6 AV |
1191 | goto failed; |
1192 | } | |
4023bfc9 | 1193 | nd->inode = inode; |
31e6b01f | 1194 | return 0; |
ef7562d5 AV |
1195 | |
1196 | failed: | |
1197 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1198 | if (!(nd->flags & LOOKUP_ROOT)) |
1199 | nd->root.mnt = NULL; | |
8b61e74f | 1200 | rcu_read_unlock(); |
ef7562d5 | 1201 | return -ECHILD; |
31e6b01f NP |
1202 | } |
1203 | ||
cc53ce53 DH |
1204 | /* |
1205 | * Follow down to the covering mount currently visible to userspace. At each | |
1206 | * point, the filesystem owning that dentry may be queried as to whether the | |
1207 | * caller is permitted to proceed or not. | |
cc53ce53 | 1208 | */ |
7cc90cc3 | 1209 | int follow_down(struct path *path) |
cc53ce53 DH |
1210 | { |
1211 | unsigned managed; | |
1212 | int ret; | |
1213 | ||
1214 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1215 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1216 | /* Allow the filesystem to manage the transit without i_mutex | |
1217 | * being held. | |
1218 | * | |
1219 | * We indicate to the filesystem if someone is trying to mount | |
1220 | * something here. This gives autofs the chance to deny anyone | |
1221 | * other than its daemon the right to mount on its | |
1222 | * superstructure. | |
1223 | * | |
1224 | * The filesystem may sleep at this point. | |
1225 | */ | |
1226 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1227 | BUG_ON(!path->dentry->d_op); | |
1228 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1229 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1230 | path->dentry, false); |
cc53ce53 DH |
1231 | if (ret < 0) |
1232 | return ret == -EISDIR ? 0 : ret; | |
1233 | } | |
1234 | ||
1235 | /* Transit to a mounted filesystem. */ | |
1236 | if (managed & DCACHE_MOUNTED) { | |
1237 | struct vfsmount *mounted = lookup_mnt(path); | |
1238 | if (!mounted) | |
1239 | break; | |
1240 | dput(path->dentry); | |
1241 | mntput(path->mnt); | |
1242 | path->mnt = mounted; | |
1243 | path->dentry = dget(mounted->mnt_root); | |
1244 | continue; | |
1245 | } | |
1246 | ||
1247 | /* Don't handle automount points here */ | |
1248 | break; | |
1249 | } | |
1250 | return 0; | |
1251 | } | |
4d359507 | 1252 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1253 | |
9875cf80 DH |
1254 | /* |
1255 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1256 | */ | |
1257 | static void follow_mount(struct path *path) | |
1258 | { | |
1259 | while (d_mountpoint(path->dentry)) { | |
1260 | struct vfsmount *mounted = lookup_mnt(path); | |
1261 | if (!mounted) | |
1262 | break; | |
1263 | dput(path->dentry); | |
1264 | mntput(path->mnt); | |
1265 | path->mnt = mounted; | |
1266 | path->dentry = dget(mounted->mnt_root); | |
1267 | } | |
1268 | } | |
1269 | ||
31e6b01f | 1270 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1271 | { |
7bd88377 AV |
1272 | if (!nd->root.mnt) |
1273 | set_root(nd); | |
e518ddb7 | 1274 | |
1da177e4 | 1275 | while(1) { |
4ac91378 | 1276 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1277 | |
2a737871 AV |
1278 | if (nd->path.dentry == nd->root.dentry && |
1279 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1280 | break; |
1281 | } | |
4ac91378 | 1282 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1283 | /* rare case of legitimate dget_parent()... */ |
1284 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1285 | dput(old); |
1286 | break; | |
1287 | } | |
3088dd70 | 1288 | if (!follow_up(&nd->path)) |
1da177e4 | 1289 | break; |
1da177e4 | 1290 | } |
79ed0226 | 1291 | follow_mount(&nd->path); |
31e6b01f | 1292 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1293 | } |
1294 | ||
baa03890 | 1295 | /* |
bad61189 MS |
1296 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1297 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1298 | * returns whether i_op->lookup is necessary. | |
1299 | * | |
1300 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1301 | */ |
bad61189 | 1302 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1303 | unsigned int flags, bool *need_lookup) |
baa03890 | 1304 | { |
baa03890 | 1305 | struct dentry *dentry; |
bad61189 | 1306 | int error; |
baa03890 | 1307 | |
bad61189 MS |
1308 | *need_lookup = false; |
1309 | dentry = d_lookup(dir, name); | |
1310 | if (dentry) { | |
39e3c955 | 1311 | if (dentry->d_flags & DCACHE_OP_REVALIDATE) { |
201f956e | 1312 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1313 | if (unlikely(error <= 0)) { |
1314 | if (error < 0) { | |
1315 | dput(dentry); | |
1316 | return ERR_PTR(error); | |
5542aa2f EB |
1317 | } else { |
1318 | d_invalidate(dentry); | |
bad61189 MS |
1319 | dput(dentry); |
1320 | dentry = NULL; | |
1321 | } | |
1322 | } | |
1323 | } | |
1324 | } | |
baa03890 | 1325 | |
bad61189 MS |
1326 | if (!dentry) { |
1327 | dentry = d_alloc(dir, name); | |
1328 | if (unlikely(!dentry)) | |
1329 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1330 | |
bad61189 | 1331 | *need_lookup = true; |
baa03890 NP |
1332 | } |
1333 | return dentry; | |
1334 | } | |
1335 | ||
44396f4b | 1336 | /* |
13a2c3be BF |
1337 | * Call i_op->lookup on the dentry. The dentry must be negative and |
1338 | * unhashed. | |
bad61189 MS |
1339 | * |
1340 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1341 | */ |
bad61189 | 1342 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1343 | unsigned int flags) |
44396f4b | 1344 | { |
44396f4b JB |
1345 | struct dentry *old; |
1346 | ||
1347 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1348 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1349 | dput(dentry); |
44396f4b | 1350 | return ERR_PTR(-ENOENT); |
e188dc02 | 1351 | } |
44396f4b | 1352 | |
72bd866a | 1353 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1354 | if (unlikely(old)) { |
1355 | dput(dentry); | |
1356 | dentry = old; | |
1357 | } | |
1358 | return dentry; | |
1359 | } | |
1360 | ||
a3255546 | 1361 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1362 | struct dentry *base, unsigned int flags) |
a3255546 | 1363 | { |
bad61189 | 1364 | bool need_lookup; |
a3255546 AV |
1365 | struct dentry *dentry; |
1366 | ||
72bd866a | 1367 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1368 | if (!need_lookup) |
1369 | return dentry; | |
a3255546 | 1370 | |
72bd866a | 1371 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1372 | } |
1373 | ||
1da177e4 LT |
1374 | /* |
1375 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1376 | * small and for now I'd prefer to have fast path as straight as possible. | |
1377 | * It _is_ time-critical. | |
1378 | */ | |
e97cdc87 | 1379 | static int lookup_fast(struct nameidata *nd, |
697f514d | 1380 | struct path *path, struct inode **inode) |
1da177e4 | 1381 | { |
4ac91378 | 1382 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1383 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1384 | int need_reval = 1; |
1385 | int status = 1; | |
9875cf80 DH |
1386 | int err; |
1387 | ||
b04f784e NP |
1388 | /* |
1389 | * Rename seqlock is not required here because in the off chance | |
1390 | * of a false negative due to a concurrent rename, we're going to | |
1391 | * do the non-racy lookup, below. | |
1392 | */ | |
31e6b01f NP |
1393 | if (nd->flags & LOOKUP_RCU) { |
1394 | unsigned seq; | |
766c4cbf | 1395 | bool negative; |
da53be12 | 1396 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5a18fff2 AV |
1397 | if (!dentry) |
1398 | goto unlazy; | |
1399 | ||
12f8ad4b LT |
1400 | /* |
1401 | * This sequence count validates that the inode matches | |
1402 | * the dentry name information from lookup. | |
1403 | */ | |
1404 | *inode = dentry->d_inode; | |
766c4cbf | 1405 | negative = d_is_negative(dentry); |
12f8ad4b LT |
1406 | if (read_seqcount_retry(&dentry->d_seq, seq)) |
1407 | return -ECHILD; | |
766c4cbf AV |
1408 | if (negative) |
1409 | return -ENOENT; | |
12f8ad4b LT |
1410 | |
1411 | /* | |
1412 | * This sequence count validates that the parent had no | |
1413 | * changes while we did the lookup of the dentry above. | |
1414 | * | |
1415 | * The memory barrier in read_seqcount_begin of child is | |
1416 | * enough, we can use __read_seqcount_retry here. | |
1417 | */ | |
31e6b01f NP |
1418 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1419 | return -ECHILD; | |
31e6b01f | 1420 | nd->seq = seq; |
5a18fff2 | 1421 | |
24643087 | 1422 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1423 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1424 | if (unlikely(status <= 0)) { |
1425 | if (status != -ECHILD) | |
1426 | need_reval = 0; | |
1427 | goto unlazy; | |
1428 | } | |
24643087 | 1429 | } |
31e6b01f NP |
1430 | path->mnt = mnt; |
1431 | path->dentry = dentry; | |
b8faf035 N |
1432 | if (likely(__follow_mount_rcu(nd, path, inode))) |
1433 | return 0; | |
5a18fff2 | 1434 | unlazy: |
19660af7 AV |
1435 | if (unlazy_walk(nd, dentry)) |
1436 | return -ECHILD; | |
5a18fff2 | 1437 | } else { |
e97cdc87 | 1438 | dentry = __d_lookup(parent, &nd->last); |
9875cf80 | 1439 | } |
5a18fff2 | 1440 | |
81e6f520 AV |
1441 | if (unlikely(!dentry)) |
1442 | goto need_lookup; | |
1443 | ||
5a18fff2 | 1444 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1445 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1446 | if (unlikely(status <= 0)) { |
1447 | if (status < 0) { | |
1448 | dput(dentry); | |
1449 | return status; | |
1450 | } | |
5542aa2f EB |
1451 | d_invalidate(dentry); |
1452 | dput(dentry); | |
1453 | goto need_lookup; | |
24643087 | 1454 | } |
697f514d | 1455 | |
766c4cbf AV |
1456 | if (unlikely(d_is_negative(dentry))) { |
1457 | dput(dentry); | |
1458 | return -ENOENT; | |
1459 | } | |
9875cf80 DH |
1460 | path->mnt = mnt; |
1461 | path->dentry = dentry; | |
1462 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1463 | if (unlikely(err < 0)) { |
1464 | path_put_conditional(path, nd); | |
9875cf80 | 1465 | return err; |
89312214 | 1466 | } |
a3fbbde7 AV |
1467 | if (err) |
1468 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1469 | *inode = path->dentry->d_inode; |
1da177e4 | 1470 | return 0; |
81e6f520 AV |
1471 | |
1472 | need_lookup: | |
697f514d MS |
1473 | return 1; |
1474 | } | |
1475 | ||
1476 | /* Fast lookup failed, do it the slow way */ | |
cc2a5271 | 1477 | static int lookup_slow(struct nameidata *nd, struct path *path) |
697f514d MS |
1478 | { |
1479 | struct dentry *dentry, *parent; | |
1480 | int err; | |
1481 | ||
1482 | parent = nd->path.dentry; | |
81e6f520 AV |
1483 | BUG_ON(nd->inode != parent->d_inode); |
1484 | ||
1485 | mutex_lock(&parent->d_inode->i_mutex); | |
cc2a5271 | 1486 | dentry = __lookup_hash(&nd->last, parent, nd->flags); |
81e6f520 AV |
1487 | mutex_unlock(&parent->d_inode->i_mutex); |
1488 | if (IS_ERR(dentry)) | |
1489 | return PTR_ERR(dentry); | |
697f514d MS |
1490 | path->mnt = nd->path.mnt; |
1491 | path->dentry = dentry; | |
1492 | err = follow_managed(path, nd->flags); | |
1493 | if (unlikely(err < 0)) { | |
1494 | path_put_conditional(path, nd); | |
1495 | return err; | |
1496 | } | |
1497 | if (err) | |
1498 | nd->flags |= LOOKUP_JUMPED; | |
1499 | return 0; | |
1da177e4 LT |
1500 | } |
1501 | ||
52094c8a AV |
1502 | static inline int may_lookup(struct nameidata *nd) |
1503 | { | |
1504 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1505 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1506 | if (err != -ECHILD) |
1507 | return err; | |
19660af7 | 1508 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1509 | return -ECHILD; |
1510 | } | |
4ad5abb3 | 1511 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1512 | } |
1513 | ||
9856fa1b AV |
1514 | static inline int handle_dots(struct nameidata *nd, int type) |
1515 | { | |
1516 | if (type == LAST_DOTDOT) { | |
1517 | if (nd->flags & LOOKUP_RCU) { | |
1518 | if (follow_dotdot_rcu(nd)) | |
1519 | return -ECHILD; | |
1520 | } else | |
1521 | follow_dotdot(nd); | |
1522 | } | |
1523 | return 0; | |
1524 | } | |
1525 | ||
951361f9 AV |
1526 | static void terminate_walk(struct nameidata *nd) |
1527 | { | |
1528 | if (!(nd->flags & LOOKUP_RCU)) { | |
1529 | path_put(&nd->path); | |
1530 | } else { | |
1531 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1532 | if (!(nd->flags & LOOKUP_ROOT)) |
1533 | nd->root.mnt = NULL; | |
8b61e74f | 1534 | rcu_read_unlock(); |
951361f9 AV |
1535 | } |
1536 | } | |
1537 | ||
3ddcd056 LT |
1538 | /* |
1539 | * Do we need to follow links? We _really_ want to be able | |
1540 | * to do this check without having to look at inode->i_op, | |
1541 | * so we keep a cache of "no, this doesn't need follow_link" | |
1542 | * for the common case. | |
1543 | */ | |
b18825a7 | 1544 | static inline int should_follow_link(struct dentry *dentry, int follow) |
3ddcd056 | 1545 | { |
b18825a7 | 1546 | return unlikely(d_is_symlink(dentry)) ? follow : 0; |
3ddcd056 LT |
1547 | } |
1548 | ||
caa85634 | 1549 | static int walk_component(struct nameidata *nd, int follow) |
ce57dfc1 | 1550 | { |
caa85634 | 1551 | struct path path; |
ce57dfc1 AV |
1552 | struct inode *inode; |
1553 | int err; | |
1554 | /* | |
1555 | * "." and ".." are special - ".." especially so because it has | |
1556 | * to be able to know about the current root directory and | |
1557 | * parent relationships. | |
1558 | */ | |
21b9b073 AV |
1559 | if (unlikely(nd->last_type != LAST_NORM)) |
1560 | return handle_dots(nd, nd->last_type); | |
caa85634 | 1561 | err = lookup_fast(nd, &path, &inode); |
ce57dfc1 | 1562 | if (unlikely(err)) { |
697f514d MS |
1563 | if (err < 0) |
1564 | goto out_err; | |
1565 | ||
caa85634 | 1566 | err = lookup_slow(nd, &path); |
697f514d MS |
1567 | if (err < 0) |
1568 | goto out_err; | |
1569 | ||
caa85634 | 1570 | inode = path.dentry->d_inode; |
766c4cbf | 1571 | err = -ENOENT; |
caa85634 | 1572 | if (d_is_negative(path.dentry)) |
766c4cbf | 1573 | goto out_path_put; |
ce57dfc1 | 1574 | } |
697f514d | 1575 | |
caa85634 | 1576 | if (should_follow_link(path.dentry, follow)) { |
19660af7 | 1577 | if (nd->flags & LOOKUP_RCU) { |
caa85634 AV |
1578 | if (unlikely(nd->path.mnt != path.mnt || |
1579 | unlazy_walk(nd, path.dentry))) { | |
697f514d MS |
1580 | err = -ECHILD; |
1581 | goto out_err; | |
19660af7 AV |
1582 | } |
1583 | } | |
caa85634 AV |
1584 | BUG_ON(inode != path.dentry->d_inode); |
1585 | nd->link = path; | |
ce57dfc1 AV |
1586 | return 1; |
1587 | } | |
caa85634 | 1588 | path_to_nameidata(&path, nd); |
ce57dfc1 AV |
1589 | nd->inode = inode; |
1590 | return 0; | |
697f514d MS |
1591 | |
1592 | out_path_put: | |
caa85634 | 1593 | path_to_nameidata(&path, nd); |
697f514d MS |
1594 | out_err: |
1595 | terminate_walk(nd); | |
1596 | return err; | |
ce57dfc1 AV |
1597 | } |
1598 | ||
bfcfaa77 LT |
1599 | /* |
1600 | * We can do the critical dentry name comparison and hashing | |
1601 | * operations one word at a time, but we are limited to: | |
1602 | * | |
1603 | * - Architectures with fast unaligned word accesses. We could | |
1604 | * do a "get_unaligned()" if this helps and is sufficiently | |
1605 | * fast. | |
1606 | * | |
bfcfaa77 LT |
1607 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1608 | * do not trap on the (extremely unlikely) case of a page | |
1609 | * crossing operation. | |
1610 | * | |
1611 | * - Furthermore, we need an efficient 64-bit compile for the | |
1612 | * 64-bit case in order to generate the "number of bytes in | |
1613 | * the final mask". Again, that could be replaced with a | |
1614 | * efficient population count instruction or similar. | |
1615 | */ | |
1616 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1617 | ||
f68e556e | 1618 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1619 | |
f68e556e | 1620 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1621 | |
1622 | static inline unsigned int fold_hash(unsigned long hash) | |
1623 | { | |
99d263d4 | 1624 | return hash_64(hash, 32); |
bfcfaa77 LT |
1625 | } |
1626 | ||
1627 | #else /* 32-bit case */ | |
1628 | ||
bfcfaa77 LT |
1629 | #define fold_hash(x) (x) |
1630 | ||
1631 | #endif | |
1632 | ||
1633 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1634 | { | |
1635 | unsigned long a, mask; | |
1636 | unsigned long hash = 0; | |
1637 | ||
1638 | for (;;) { | |
e419b4cc | 1639 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1640 | if (len < sizeof(unsigned long)) |
1641 | break; | |
1642 | hash += a; | |
f132c5be | 1643 | hash *= 9; |
bfcfaa77 LT |
1644 | name += sizeof(unsigned long); |
1645 | len -= sizeof(unsigned long); | |
1646 | if (!len) | |
1647 | goto done; | |
1648 | } | |
a5c21dce | 1649 | mask = bytemask_from_count(len); |
bfcfaa77 LT |
1650 | hash += mask & a; |
1651 | done: | |
1652 | return fold_hash(hash); | |
1653 | } | |
1654 | EXPORT_SYMBOL(full_name_hash); | |
1655 | ||
bfcfaa77 LT |
1656 | /* |
1657 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1658 | * return the "hash_len" as the result. |
bfcfaa77 | 1659 | */ |
d6bb3e90 | 1660 | static inline u64 hash_name(const char *name) |
bfcfaa77 | 1661 | { |
36126f8f LT |
1662 | unsigned long a, b, adata, bdata, mask, hash, len; |
1663 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1664 | |
1665 | hash = a = 0; | |
1666 | len = -sizeof(unsigned long); | |
1667 | do { | |
1668 | hash = (hash + a) * 9; | |
1669 | len += sizeof(unsigned long); | |
e419b4cc | 1670 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1671 | b = a ^ REPEAT_BYTE('/'); |
1672 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1673 | ||
1674 | adata = prep_zero_mask(a, adata, &constants); | |
1675 | bdata = prep_zero_mask(b, bdata, &constants); | |
1676 | ||
1677 | mask = create_zero_mask(adata | bdata); | |
1678 | ||
1679 | hash += a & zero_bytemask(mask); | |
9226b5b4 | 1680 | len += find_zero(mask); |
d6bb3e90 | 1681 | return hashlen_create(fold_hash(hash), len); |
bfcfaa77 LT |
1682 | } |
1683 | ||
1684 | #else | |
1685 | ||
0145acc2 LT |
1686 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1687 | { | |
1688 | unsigned long hash = init_name_hash(); | |
1689 | while (len--) | |
1690 | hash = partial_name_hash(*name++, hash); | |
1691 | return end_name_hash(hash); | |
1692 | } | |
ae942ae7 | 1693 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1694 | |
200e9ef7 LT |
1695 | /* |
1696 | * We know there's a real path component here of at least | |
1697 | * one character. | |
1698 | */ | |
d6bb3e90 | 1699 | static inline u64 hash_name(const char *name) |
200e9ef7 LT |
1700 | { |
1701 | unsigned long hash = init_name_hash(); | |
1702 | unsigned long len = 0, c; | |
1703 | ||
1704 | c = (unsigned char)*name; | |
1705 | do { | |
1706 | len++; | |
1707 | hash = partial_name_hash(c, hash); | |
1708 | c = (unsigned char)name[len]; | |
1709 | } while (c && c != '/'); | |
d6bb3e90 | 1710 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
1711 | } |
1712 | ||
bfcfaa77 LT |
1713 | #endif |
1714 | ||
1da177e4 LT |
1715 | /* |
1716 | * Name resolution. | |
ea3834d9 PM |
1717 | * This is the basic name resolution function, turning a pathname into |
1718 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1719 | * |
ea3834d9 PM |
1720 | * Returns 0 and nd will have valid dentry and mnt on success. |
1721 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1722 | */ |
6de88d72 | 1723 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 1724 | { |
1da177e4 | 1725 | int err; |
32cd7468 | 1726 | |
1da177e4 LT |
1727 | while (*name=='/') |
1728 | name++; | |
1729 | if (!*name) | |
9e18f10a | 1730 | return 0; |
1da177e4 | 1731 | |
1da177e4 LT |
1732 | /* At this point we know we have a real path component. */ |
1733 | for(;;) { | |
d6bb3e90 | 1734 | u64 hash_len; |
fe479a58 | 1735 | int type; |
1da177e4 | 1736 | |
52094c8a | 1737 | err = may_lookup(nd); |
1da177e4 LT |
1738 | if (err) |
1739 | break; | |
1740 | ||
d6bb3e90 | 1741 | hash_len = hash_name(name); |
1da177e4 | 1742 | |
fe479a58 | 1743 | type = LAST_NORM; |
d6bb3e90 | 1744 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 1745 | case 2: |
200e9ef7 | 1746 | if (name[1] == '.') { |
fe479a58 | 1747 | type = LAST_DOTDOT; |
16c2cd71 AV |
1748 | nd->flags |= LOOKUP_JUMPED; |
1749 | } | |
fe479a58 AV |
1750 | break; |
1751 | case 1: | |
1752 | type = LAST_DOT; | |
1753 | } | |
5a202bcd AV |
1754 | if (likely(type == LAST_NORM)) { |
1755 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1756 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 1757 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 1758 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 1759 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd AV |
1760 | if (err < 0) |
1761 | break; | |
d6bb3e90 LT |
1762 | hash_len = this.hash_len; |
1763 | name = this.name; | |
5a202bcd AV |
1764 | } |
1765 | } | |
fe479a58 | 1766 | |
d6bb3e90 LT |
1767 | nd->last.hash_len = hash_len; |
1768 | nd->last.name = name; | |
5f4a6a69 AV |
1769 | nd->last_type = type; |
1770 | ||
d6bb3e90 LT |
1771 | name += hashlen_len(hash_len); |
1772 | if (!*name) | |
bdf6cbf1 | 1773 | goto OK; |
200e9ef7 LT |
1774 | /* |
1775 | * If it wasn't NUL, we know it was '/'. Skip that | |
1776 | * slash, and continue until no more slashes. | |
1777 | */ | |
1778 | do { | |
d6bb3e90 LT |
1779 | name++; |
1780 | } while (unlikely(*name == '/')); | |
1781 | if (!*name) | |
bdf6cbf1 | 1782 | goto OK; |
5f4a6a69 | 1783 | |
caa85634 | 1784 | err = walk_component(nd, LOOKUP_FOLLOW); |
bb8603f8 | 1785 | Walked: |
ce57dfc1 | 1786 | if (err < 0) |
bdf6cbf1 | 1787 | goto Err; |
1da177e4 | 1788 | |
ce57dfc1 | 1789 | if (err) { |
d40bcc09 AV |
1790 | const char *s; |
1791 | ||
5a460275 AV |
1792 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1793 | path_put_conditional(&nd->link, nd); | |
1794 | path_put(&nd->path); | |
bdf6cbf1 AV |
1795 | err = -ELOOP; |
1796 | goto Err; | |
5a460275 AV |
1797 | } |
1798 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
1799 | ||
1800 | nd->depth++; | |
1801 | current->link_count++; | |
1802 | ||
3b2e7f75 | 1803 | s = get_link(nd); |
5a460275 | 1804 | |
d40bcc09 AV |
1805 | if (unlikely(IS_ERR(s))) { |
1806 | err = PTR_ERR(s); | |
1807 | current->link_count--; | |
1808 | nd->depth--; | |
bdf6cbf1 | 1809 | goto Err; |
d40bcc09 AV |
1810 | } |
1811 | err = 0; | |
1812 | if (unlikely(!s)) { | |
1813 | /* jumped */ | |
b9ff4429 | 1814 | put_link(nd); |
48c8b0c5 AV |
1815 | current->link_count--; |
1816 | nd->depth--; | |
d40bcc09 AV |
1817 | } else { |
1818 | if (*s == '/') { | |
1819 | if (!nd->root.mnt) | |
1820 | set_root(nd); | |
1821 | path_put(&nd->path); | |
1822 | nd->path = nd->root; | |
1823 | path_get(&nd->root); | |
1824 | nd->flags |= LOOKUP_JUMPED; | |
9e18f10a AV |
1825 | while (unlikely(*++s == '/')) |
1826 | ; | |
172a39a0 | 1827 | } |
d40bcc09 | 1828 | nd->inode = nd->path.dentry->d_inode; |
3b2e7f75 | 1829 | nd->stack[nd->depth].name = name; |
9e18f10a AV |
1830 | if (!*s) |
1831 | goto OK; | |
32cd7468 | 1832 | name = s; |
9e18f10a | 1833 | continue; |
d40bcc09 | 1834 | } |
31e6b01f | 1835 | } |
44b1d530 | 1836 | if (!d_can_lookup(nd->path.dentry)) { |
caa85634 | 1837 | err = -ENOTDIR; |
5f4a6a69 AV |
1838 | break; |
1839 | } | |
1da177e4 | 1840 | } |
951361f9 | 1841 | terminate_walk(nd); |
bdf6cbf1 | 1842 | Err: |
07681481 | 1843 | while (unlikely(nd->depth)) { |
b9ff4429 | 1844 | put_link(nd); |
07681481 AV |
1845 | current->link_count--; |
1846 | nd->depth--; | |
07681481 AV |
1847 | } |
1848 | return err; | |
bdf6cbf1 | 1849 | OK: |
07681481 | 1850 | if (unlikely(nd->depth)) { |
3b2e7f75 | 1851 | name = nd->stack[nd->depth].name; |
07681481 | 1852 | err = walk_component(nd, LOOKUP_FOLLOW); |
b9ff4429 | 1853 | put_link(nd); |
07681481 AV |
1854 | current->link_count--; |
1855 | nd->depth--; | |
07681481 AV |
1856 | goto Walked; |
1857 | } | |
1858 | return 0; | |
1da177e4 LT |
1859 | } |
1860 | ||
6e8a1f87 | 1861 | static int path_init(int dfd, const struct filename *name, unsigned int flags, |
5e53084d | 1862 | struct nameidata *nd) |
31e6b01f NP |
1863 | { |
1864 | int retval = 0; | |
fd2f7cb5 | 1865 | const char *s = name->name; |
31e6b01f NP |
1866 | |
1867 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
980f3ea2 | 1868 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 1869 | nd->depth = 0; |
5e53084d | 1870 | nd->base = NULL; |
5b6ca027 | 1871 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
1872 | struct dentry *root = nd->root.dentry; |
1873 | struct inode *inode = root->d_inode; | |
fd2f7cb5 | 1874 | if (*s) { |
44b1d530 | 1875 | if (!d_can_lookup(root)) |
73d049a4 AV |
1876 | return -ENOTDIR; |
1877 | retval = inode_permission(inode, MAY_EXEC); | |
1878 | if (retval) | |
1879 | return retval; | |
1880 | } | |
5b6ca027 AV |
1881 | nd->path = nd->root; |
1882 | nd->inode = inode; | |
1883 | if (flags & LOOKUP_RCU) { | |
8b61e74f | 1884 | rcu_read_lock(); |
5b6ca027 | 1885 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
48a066e7 | 1886 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
1887 | } else { |
1888 | path_get(&nd->path); | |
1889 | } | |
d465887f | 1890 | goto done; |
5b6ca027 AV |
1891 | } |
1892 | ||
31e6b01f | 1893 | nd->root.mnt = NULL; |
31e6b01f | 1894 | |
48a066e7 | 1895 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 1896 | if (*s == '/') { |
e41f7d4e | 1897 | if (flags & LOOKUP_RCU) { |
8b61e74f | 1898 | rcu_read_lock(); |
7bd88377 | 1899 | nd->seq = set_root_rcu(nd); |
e41f7d4e AV |
1900 | } else { |
1901 | set_root(nd); | |
1902 | path_get(&nd->root); | |
1903 | } | |
1904 | nd->path = nd->root; | |
31e6b01f | 1905 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1906 | if (flags & LOOKUP_RCU) { |
1907 | struct fs_struct *fs = current->fs; | |
1908 | unsigned seq; | |
31e6b01f | 1909 | |
8b61e74f | 1910 | rcu_read_lock(); |
c28cc364 | 1911 | |
e41f7d4e AV |
1912 | do { |
1913 | seq = read_seqcount_begin(&fs->seq); | |
1914 | nd->path = fs->pwd; | |
1915 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1916 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1917 | } else { | |
1918 | get_fs_pwd(current->fs, &nd->path); | |
1919 | } | |
31e6b01f | 1920 | } else { |
582aa64a | 1921 | /* Caller must check execute permissions on the starting path component */ |
2903ff01 | 1922 | struct fd f = fdget_raw(dfd); |
31e6b01f NP |
1923 | struct dentry *dentry; |
1924 | ||
2903ff01 AV |
1925 | if (!f.file) |
1926 | return -EBADF; | |
31e6b01f | 1927 | |
2903ff01 | 1928 | dentry = f.file->f_path.dentry; |
31e6b01f | 1929 | |
fd2f7cb5 | 1930 | if (*s) { |
44b1d530 | 1931 | if (!d_can_lookup(dentry)) { |
2903ff01 AV |
1932 | fdput(f); |
1933 | return -ENOTDIR; | |
1934 | } | |
f52e0c11 | 1935 | } |
31e6b01f | 1936 | |
2903ff01 | 1937 | nd->path = f.file->f_path; |
e41f7d4e | 1938 | if (flags & LOOKUP_RCU) { |
9c225f26 | 1939 | if (f.flags & FDPUT_FPUT) |
5e53084d | 1940 | nd->base = f.file; |
e41f7d4e | 1941 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
8b61e74f | 1942 | rcu_read_lock(); |
e41f7d4e | 1943 | } else { |
2903ff01 AV |
1944 | path_get(&nd->path); |
1945 | fdput(f); | |
e41f7d4e | 1946 | } |
31e6b01f | 1947 | } |
31e6b01f | 1948 | |
31e6b01f | 1949 | nd->inode = nd->path.dentry->d_inode; |
4023bfc9 | 1950 | if (!(flags & LOOKUP_RCU)) |
d465887f | 1951 | goto done; |
4023bfc9 | 1952 | if (likely(!read_seqcount_retry(&nd->path.dentry->d_seq, nd->seq))) |
d465887f | 1953 | goto done; |
4023bfc9 AV |
1954 | if (!(nd->flags & LOOKUP_ROOT)) |
1955 | nd->root.mnt = NULL; | |
1956 | rcu_read_unlock(); | |
1957 | return -ECHILD; | |
d465887f AV |
1958 | done: |
1959 | current->total_link_count = 0; | |
fd2f7cb5 | 1960 | return link_path_walk(s, nd); |
9b4a9b14 AV |
1961 | } |
1962 | ||
893b7775 AV |
1963 | static void path_cleanup(struct nameidata *nd) |
1964 | { | |
1965 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
1966 | path_put(&nd->root); | |
1967 | nd->root.mnt = NULL; | |
1968 | } | |
1969 | if (unlikely(nd->base)) | |
1970 | fput(nd->base); | |
9b4a9b14 AV |
1971 | } |
1972 | ||
1d8e03d3 | 1973 | static int trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
1974 | { |
1975 | const char *s; | |
1d8e03d3 | 1976 | int error = may_follow_link(&nd->link, nd); |
95fa25d9 AV |
1977 | if (unlikely(error)) |
1978 | return error; | |
1979 | nd->flags |= LOOKUP_PARENT; | |
3b2e7f75 | 1980 | s = get_link(nd); |
95fa25d9 AV |
1981 | if (unlikely(IS_ERR(s))) |
1982 | return PTR_ERR(s); | |
1983 | if (unlikely(!s)) | |
1984 | return 0; | |
1985 | if (*s == '/') { | |
1986 | if (!nd->root.mnt) | |
1987 | set_root(nd); | |
1988 | path_put(&nd->path); | |
1989 | nd->path = nd->root; | |
1990 | path_get(&nd->root); | |
1991 | nd->flags |= LOOKUP_JUMPED; | |
1992 | } | |
1993 | nd->inode = nd->path.dentry->d_inode; | |
1994 | error = link_path_walk(s, nd); | |
1995 | if (unlikely(error)) | |
b9ff4429 | 1996 | put_link(nd); |
95fa25d9 AV |
1997 | return error; |
1998 | } | |
1999 | ||
caa85634 | 2000 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2001 | { |
2002 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2003 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2004 | ||
2005 | nd->flags &= ~LOOKUP_PARENT; | |
caa85634 | 2006 | return walk_component(nd, nd->flags & LOOKUP_FOLLOW); |
bd92d7fe AV |
2007 | } |
2008 | ||
9b4a9b14 | 2009 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
5eb6b495 | 2010 | static int path_lookupat(int dfd, const struct filename *name, |
9b4a9b14 AV |
2011 | unsigned int flags, struct nameidata *nd) |
2012 | { | |
bd92d7fe | 2013 | int err; |
31e6b01f NP |
2014 | |
2015 | /* | |
2016 | * Path walking is largely split up into 2 different synchronisation | |
2017 | * schemes, rcu-walk and ref-walk (explained in | |
2018 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
2019 | * path walk code, but some things particularly setup, cleanup, and | |
2020 | * following mounts are sufficiently divergent that functions are | |
2021 | * duplicated. Typically there is a function foo(), and its RCU | |
2022 | * analogue, foo_rcu(). | |
2023 | * | |
2024 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
2025 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
2026 | * be handled by restarting a traditional ref-walk (which will always | |
2027 | * be able to complete). | |
2028 | */ | |
6e8a1f87 | 2029 | err = path_init(dfd, name, flags, nd); |
bd92d7fe | 2030 | if (!err && !(flags & LOOKUP_PARENT)) { |
caa85634 | 2031 | err = lookup_last(nd); |
bd92d7fe | 2032 | while (err > 0) { |
1d8e03d3 | 2033 | err = trailing_symlink(nd); |
6d7b5aae AV |
2034 | if (err) |
2035 | break; | |
caa85634 | 2036 | err = lookup_last(nd); |
b9ff4429 | 2037 | put_link(nd); |
bd92d7fe AV |
2038 | } |
2039 | } | |
ee0827cd | 2040 | |
9f1fafee AV |
2041 | if (!err) |
2042 | err = complete_walk(nd); | |
bd92d7fe AV |
2043 | |
2044 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
44b1d530 | 2045 | if (!d_can_lookup(nd->path.dentry)) { |
bd92d7fe | 2046 | path_put(&nd->path); |
bd23a539 | 2047 | err = -ENOTDIR; |
bd92d7fe AV |
2048 | } |
2049 | } | |
16c2cd71 | 2050 | |
893b7775 | 2051 | path_cleanup(nd); |
bd92d7fe | 2052 | return err; |
ee0827cd | 2053 | } |
31e6b01f | 2054 | |
873f1eed | 2055 | static int filename_lookup(int dfd, struct filename *name, |
ee0827cd AV |
2056 | unsigned int flags, struct nameidata *nd) |
2057 | { | |
5eb6b495 | 2058 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); |
ee0827cd | 2059 | if (unlikely(retval == -ECHILD)) |
5eb6b495 | 2060 | retval = path_lookupat(dfd, name, flags, nd); |
ee0827cd | 2061 | if (unlikely(retval == -ESTALE)) |
5eb6b495 | 2062 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); |
31e6b01f | 2063 | |
f78570dd | 2064 | if (likely(!retval)) |
adb5c247 | 2065 | audit_inode(name, nd->path.dentry, flags & LOOKUP_PARENT); |
170aa3d0 | 2066 | return retval; |
1da177e4 LT |
2067 | } |
2068 | ||
79714f72 AV |
2069 | /* does lookup, returns the object with parent locked */ |
2070 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2071 | { |
51689104 | 2072 | struct filename *filename = getname_kernel(name); |
79714f72 AV |
2073 | struct nameidata nd; |
2074 | struct dentry *d; | |
51689104 PM |
2075 | int err; |
2076 | ||
2077 | if (IS_ERR(filename)) | |
2078 | return ERR_CAST(filename); | |
2079 | ||
2080 | err = filename_lookup(AT_FDCWD, filename, LOOKUP_PARENT, &nd); | |
2081 | if (err) { | |
2082 | d = ERR_PTR(err); | |
2083 | goto out; | |
2084 | } | |
79714f72 AV |
2085 | if (nd.last_type != LAST_NORM) { |
2086 | path_put(&nd.path); | |
51689104 PM |
2087 | d = ERR_PTR(-EINVAL); |
2088 | goto out; | |
79714f72 AV |
2089 | } |
2090 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 2091 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
2092 | if (IS_ERR(d)) { |
2093 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
2094 | path_put(&nd.path); | |
51689104 | 2095 | goto out; |
79714f72 AV |
2096 | } |
2097 | *path = nd.path; | |
51689104 PM |
2098 | out: |
2099 | putname(filename); | |
79714f72 | 2100 | return d; |
5590ff0d UD |
2101 | } |
2102 | ||
d1811465 AV |
2103 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2104 | { | |
2105 | struct nameidata nd; | |
74eb8cc5 AV |
2106 | struct filename *filename = getname_kernel(name); |
2107 | int res = PTR_ERR(filename); | |
2108 | ||
2109 | if (!IS_ERR(filename)) { | |
2110 | res = filename_lookup(AT_FDCWD, filename, flags, &nd); | |
2111 | putname(filename); | |
2112 | if (!res) | |
2113 | *path = nd.path; | |
2114 | } | |
d1811465 AV |
2115 | return res; |
2116 | } | |
4d359507 | 2117 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2118 | |
16f18200 JJS |
2119 | /** |
2120 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2121 | * @dentry: pointer to dentry of the base directory | |
2122 | * @mnt: pointer to vfs mount of the base directory | |
2123 | * @name: pointer to file name | |
2124 | * @flags: lookup flags | |
e0a01249 | 2125 | * @path: pointer to struct path to fill |
16f18200 JJS |
2126 | */ |
2127 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2128 | const char *name, unsigned int flags, | |
e0a01249 | 2129 | struct path *path) |
16f18200 | 2130 | { |
74eb8cc5 AV |
2131 | struct filename *filename = getname_kernel(name); |
2132 | int err = PTR_ERR(filename); | |
2133 | ||
e0a01249 | 2134 | BUG_ON(flags & LOOKUP_PARENT); |
74eb8cc5 AV |
2135 | |
2136 | /* the first argument of filename_lookup() is ignored with LOOKUP_ROOT */ | |
2137 | if (!IS_ERR(filename)) { | |
2138 | struct nameidata nd; | |
2139 | nd.root.dentry = dentry; | |
2140 | nd.root.mnt = mnt; | |
2141 | err = filename_lookup(AT_FDCWD, filename, | |
2142 | flags | LOOKUP_ROOT, &nd); | |
2143 | if (!err) | |
2144 | *path = nd.path; | |
2145 | putname(filename); | |
2146 | } | |
e0a01249 | 2147 | return err; |
16f18200 | 2148 | } |
4d359507 | 2149 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2150 | |
eead1911 | 2151 | /** |
a6b91919 | 2152 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2153 | * @name: pathname component to lookup |
2154 | * @base: base directory to lookup from | |
2155 | * @len: maximum length @len should be interpreted to | |
2156 | * | |
a6b91919 | 2157 | * Note that this routine is purely a helper for filesystem usage and should |
9e7543e9 | 2158 | * not be called by generic code. |
eead1911 | 2159 | */ |
057f6c01 JM |
2160 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2161 | { | |
057f6c01 | 2162 | struct qstr this; |
6a96ba54 | 2163 | unsigned int c; |
cda309de | 2164 | int err; |
057f6c01 | 2165 | |
2f9092e1 DW |
2166 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
2167 | ||
6a96ba54 AV |
2168 | this.name = name; |
2169 | this.len = len; | |
0145acc2 | 2170 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2171 | if (!len) |
2172 | return ERR_PTR(-EACCES); | |
2173 | ||
21d8a15a AV |
2174 | if (unlikely(name[0] == '.')) { |
2175 | if (len < 2 || (len == 2 && name[1] == '.')) | |
2176 | return ERR_PTR(-EACCES); | |
2177 | } | |
2178 | ||
6a96ba54 AV |
2179 | while (len--) { |
2180 | c = *(const unsigned char *)name++; | |
2181 | if (c == '/' || c == '\0') | |
2182 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2183 | } |
5a202bcd AV |
2184 | /* |
2185 | * See if the low-level filesystem might want | |
2186 | * to use its own hash.. | |
2187 | */ | |
2188 | if (base->d_flags & DCACHE_OP_HASH) { | |
da53be12 | 2189 | int err = base->d_op->d_hash(base, &this); |
5a202bcd AV |
2190 | if (err < 0) |
2191 | return ERR_PTR(err); | |
2192 | } | |
eead1911 | 2193 | |
cda309de MS |
2194 | err = inode_permission(base->d_inode, MAY_EXEC); |
2195 | if (err) | |
2196 | return ERR_PTR(err); | |
2197 | ||
72bd866a | 2198 | return __lookup_hash(&this, base, 0); |
057f6c01 | 2199 | } |
4d359507 | 2200 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2201 | |
1fa1e7f6 AW |
2202 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2203 | struct path *path, int *empty) | |
1da177e4 | 2204 | { |
2d8f3038 | 2205 | struct nameidata nd; |
91a27b2a | 2206 | struct filename *tmp = getname_flags(name, flags, empty); |
1da177e4 | 2207 | int err = PTR_ERR(tmp); |
1da177e4 | 2208 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
2209 | |
2210 | BUG_ON(flags & LOOKUP_PARENT); | |
2211 | ||
873f1eed | 2212 | err = filename_lookup(dfd, tmp, flags, &nd); |
1da177e4 | 2213 | putname(tmp); |
2d8f3038 AV |
2214 | if (!err) |
2215 | *path = nd.path; | |
1da177e4 LT |
2216 | } |
2217 | return err; | |
2218 | } | |
2219 | ||
1fa1e7f6 AW |
2220 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
2221 | struct path *path) | |
2222 | { | |
f7493e5d | 2223 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 | 2224 | } |
4d359507 | 2225 | EXPORT_SYMBOL(user_path_at); |
1fa1e7f6 | 2226 | |
873f1eed JL |
2227 | /* |
2228 | * NB: most callers don't do anything directly with the reference to the | |
2229 | * to struct filename, but the nd->last pointer points into the name string | |
2230 | * allocated by getname. So we must hold the reference to it until all | |
2231 | * path-walking is complete. | |
2232 | */ | |
91a27b2a | 2233 | static struct filename * |
f5beed75 AV |
2234 | user_path_parent(int dfd, const char __user *path, |
2235 | struct path *parent, | |
2236 | struct qstr *last, | |
2237 | int *type, | |
9e790bd6 | 2238 | unsigned int flags) |
2ad94ae6 | 2239 | { |
f5beed75 | 2240 | struct nameidata nd; |
91a27b2a | 2241 | struct filename *s = getname(path); |
2ad94ae6 AV |
2242 | int error; |
2243 | ||
9e790bd6 JL |
2244 | /* only LOOKUP_REVAL is allowed in extra flags */ |
2245 | flags &= LOOKUP_REVAL; | |
2246 | ||
2ad94ae6 | 2247 | if (IS_ERR(s)) |
91a27b2a | 2248 | return s; |
2ad94ae6 | 2249 | |
f5beed75 | 2250 | error = filename_lookup(dfd, s, flags | LOOKUP_PARENT, &nd); |
91a27b2a | 2251 | if (error) { |
2ad94ae6 | 2252 | putname(s); |
91a27b2a JL |
2253 | return ERR_PTR(error); |
2254 | } | |
f5beed75 AV |
2255 | *parent = nd.path; |
2256 | *last = nd.last; | |
2257 | *type = nd.last_type; | |
2ad94ae6 | 2258 | |
91a27b2a | 2259 | return s; |
2ad94ae6 AV |
2260 | } |
2261 | ||
8033426e | 2262 | /** |
197df04c | 2263 | * mountpoint_last - look up last component for umount |
8033426e JL |
2264 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
2265 | * @path: pointer to container for result | |
2266 | * | |
2267 | * This is a special lookup_last function just for umount. In this case, we | |
2268 | * need to resolve the path without doing any revalidation. | |
2269 | * | |
2270 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2271 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2272 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2273 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2274 | * bogus and it doesn't exist. | |
2275 | * | |
2276 | * Returns: | |
2277 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
2278 | * lookup found a negative dentry. The nd->path reference will also be | |
2279 | * put in this case. | |
2280 | * | |
2281 | * 0: if we successfully resolved nd->path and found it to not to be a | |
2282 | * symlink that needs to be followed. "path" will also be populated. | |
2283 | * The nd->path reference will also be put. | |
2284 | * | |
2285 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
2286 | * that needs to be followed. "path" will be populated with the path | |
2287 | * to the link, and nd->path will *not* be put. | |
2288 | */ | |
2289 | static int | |
197df04c | 2290 | mountpoint_last(struct nameidata *nd, struct path *path) |
8033426e JL |
2291 | { |
2292 | int error = 0; | |
2293 | struct dentry *dentry; | |
2294 | struct dentry *dir = nd->path.dentry; | |
2295 | ||
35759521 AV |
2296 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2297 | if (nd->flags & LOOKUP_RCU) { | |
2298 | if (unlazy_walk(nd, NULL)) { | |
2299 | error = -ECHILD; | |
2300 | goto out; | |
2301 | } | |
8033426e JL |
2302 | } |
2303 | ||
2304 | nd->flags &= ~LOOKUP_PARENT; | |
2305 | ||
2306 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2307 | error = handle_dots(nd, nd->last_type); | |
35759521 AV |
2308 | if (error) |
2309 | goto out; | |
2310 | dentry = dget(nd->path.dentry); | |
2311 | goto done; | |
8033426e JL |
2312 | } |
2313 | ||
2314 | mutex_lock(&dir->d_inode->i_mutex); | |
2315 | dentry = d_lookup(dir, &nd->last); | |
2316 | if (!dentry) { | |
2317 | /* | |
2318 | * No cached dentry. Mounted dentries are pinned in the cache, | |
2319 | * so that means that this dentry is probably a symlink or the | |
2320 | * path doesn't actually point to a mounted dentry. | |
2321 | */ | |
2322 | dentry = d_alloc(dir, &nd->last); | |
2323 | if (!dentry) { | |
2324 | error = -ENOMEM; | |
bcceeeba | 2325 | mutex_unlock(&dir->d_inode->i_mutex); |
35759521 | 2326 | goto out; |
8033426e | 2327 | } |
35759521 AV |
2328 | dentry = lookup_real(dir->d_inode, dentry, nd->flags); |
2329 | error = PTR_ERR(dentry); | |
bcceeeba DJ |
2330 | if (IS_ERR(dentry)) { |
2331 | mutex_unlock(&dir->d_inode->i_mutex); | |
35759521 | 2332 | goto out; |
bcceeeba | 2333 | } |
8033426e JL |
2334 | } |
2335 | mutex_unlock(&dir->d_inode->i_mutex); | |
2336 | ||
35759521 | 2337 | done: |
698934df | 2338 | if (d_is_negative(dentry)) { |
35759521 AV |
2339 | error = -ENOENT; |
2340 | dput(dentry); | |
2341 | goto out; | |
8033426e | 2342 | } |
35759521 | 2343 | path->dentry = dentry; |
295dc39d | 2344 | path->mnt = nd->path.mnt; |
caa85634 AV |
2345 | if (should_follow_link(dentry, nd->flags & LOOKUP_FOLLOW)) { |
2346 | nd->link = *path; | |
35759521 | 2347 | return 1; |
caa85634 | 2348 | } |
295dc39d | 2349 | mntget(path->mnt); |
35759521 AV |
2350 | follow_mount(path); |
2351 | error = 0; | |
2352 | out: | |
8033426e JL |
2353 | terminate_walk(nd); |
2354 | return error; | |
2355 | } | |
2356 | ||
2357 | /** | |
197df04c | 2358 | * path_mountpoint - look up a path to be umounted |
8033426e JL |
2359 | * @dfd: directory file descriptor to start walk from |
2360 | * @name: full pathname to walk | |
606d6fe3 | 2361 | * @path: pointer to container for result |
8033426e | 2362 | * @flags: lookup flags |
8033426e JL |
2363 | * |
2364 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2365 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2366 | */ |
2367 | static int | |
668696dc | 2368 | path_mountpoint(int dfd, const struct filename *name, struct path *path, |
46afd6f6 | 2369 | struct nameidata *nd, unsigned int flags) |
8033426e | 2370 | { |
46afd6f6 | 2371 | int err = path_init(dfd, name, flags, nd); |
8033426e | 2372 | if (unlikely(err)) |
115cbfdc | 2373 | goto out; |
8033426e | 2374 | |
46afd6f6 | 2375 | err = mountpoint_last(nd, path); |
8033426e | 2376 | while (err > 0) { |
1d8e03d3 | 2377 | err = trailing_symlink(nd); |
8033426e JL |
2378 | if (err) |
2379 | break; | |
46afd6f6 | 2380 | err = mountpoint_last(nd, path); |
b9ff4429 | 2381 | put_link(nd); |
8033426e JL |
2382 | } |
2383 | out: | |
46afd6f6 | 2384 | path_cleanup(nd); |
8033426e JL |
2385 | return err; |
2386 | } | |
2387 | ||
2d864651 | 2388 | static int |
668696dc | 2389 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2390 | unsigned int flags) |
2391 | { | |
46afd6f6 | 2392 | struct nameidata nd; |
cbaab2db | 2393 | int error; |
668696dc AV |
2394 | if (IS_ERR(name)) |
2395 | return PTR_ERR(name); | |
46afd6f6 | 2396 | error = path_mountpoint(dfd, name, path, &nd, flags | LOOKUP_RCU); |
2d864651 | 2397 | if (unlikely(error == -ECHILD)) |
46afd6f6 | 2398 | error = path_mountpoint(dfd, name, path, &nd, flags); |
2d864651 | 2399 | if (unlikely(error == -ESTALE)) |
46afd6f6 | 2400 | error = path_mountpoint(dfd, name, path, &nd, flags | LOOKUP_REVAL); |
2d864651 | 2401 | if (likely(!error)) |
668696dc AV |
2402 | audit_inode(name, path->dentry, 0); |
2403 | putname(name); | |
2d864651 AV |
2404 | return error; |
2405 | } | |
2406 | ||
8033426e | 2407 | /** |
197df04c | 2408 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2409 | * @dfd: directory file descriptor |
2410 | * @name: pathname from userland | |
2411 | * @flags: lookup flags | |
2412 | * @path: pointer to container to hold result | |
2413 | * | |
2414 | * A umount is a special case for path walking. We're not actually interested | |
2415 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2416 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2417 | * and avoid revalidating the last component. | |
2418 | * | |
2419 | * Returns 0 and populates "path" on success. | |
2420 | */ | |
2421 | int | |
197df04c | 2422 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2423 | struct path *path) |
2424 | { | |
cbaab2db | 2425 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2426 | } |
2427 | ||
2d864651 AV |
2428 | int |
2429 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2430 | unsigned int flags) | |
2431 | { | |
cbaab2db | 2432 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2433 | } |
2434 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2435 | ||
cbdf35bc | 2436 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2437 | { |
8e96e3b7 | 2438 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2439 | |
8e96e3b7 | 2440 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2441 | return 0; |
8e96e3b7 | 2442 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2443 | return 0; |
23adbe12 | 2444 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2445 | } |
cbdf35bc | 2446 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2447 | |
2448 | /* | |
2449 | * Check whether we can remove a link victim from directory dir, check | |
2450 | * whether the type of victim is right. | |
2451 | * 1. We can't do it if dir is read-only (done in permission()) | |
2452 | * 2. We should have write and exec permissions on dir | |
2453 | * 3. We can't remove anything from append-only dir | |
2454 | * 4. We can't do anything with immutable dir (done in permission()) | |
2455 | * 5. If the sticky bit on dir is set we should either | |
2456 | * a. be owner of dir, or | |
2457 | * b. be owner of victim, or | |
2458 | * c. have CAP_FOWNER capability | |
2459 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2460 | * links pointing to it. | |
2461 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2462 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2463 | * 9. We can't remove a root or mountpoint. | |
2464 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2465 | * nfs_async_unlink(). | |
2466 | */ | |
b18825a7 | 2467 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2468 | { |
b18825a7 | 2469 | struct inode *inode = victim->d_inode; |
1da177e4 LT |
2470 | int error; |
2471 | ||
b18825a7 | 2472 | if (d_is_negative(victim)) |
1da177e4 | 2473 | return -ENOENT; |
b18825a7 | 2474 | BUG_ON(!inode); |
1da177e4 LT |
2475 | |
2476 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 2477 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2478 | |
f419a2e3 | 2479 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2480 | if (error) |
2481 | return error; | |
2482 | if (IS_APPEND(dir)) | |
2483 | return -EPERM; | |
b18825a7 DH |
2484 | |
2485 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
2486 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode)) | |
1da177e4 LT |
2487 | return -EPERM; |
2488 | if (isdir) { | |
44b1d530 | 2489 | if (!d_is_dir(victim)) |
1da177e4 LT |
2490 | return -ENOTDIR; |
2491 | if (IS_ROOT(victim)) | |
2492 | return -EBUSY; | |
44b1d530 | 2493 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2494 | return -EISDIR; |
2495 | if (IS_DEADDIR(dir)) | |
2496 | return -ENOENT; | |
2497 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2498 | return -EBUSY; | |
2499 | return 0; | |
2500 | } | |
2501 | ||
2502 | /* Check whether we can create an object with dentry child in directory | |
2503 | * dir. | |
2504 | * 1. We can't do it if child already exists (open has special treatment for | |
2505 | * this case, but since we are inlined it's OK) | |
2506 | * 2. We can't do it if dir is read-only (done in permission()) | |
2507 | * 3. We should have write and exec permissions on dir | |
2508 | * 4. We can't do it if dir is immutable (done in permission()) | |
2509 | */ | |
a95164d9 | 2510 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2511 | { |
14e972b4 | 2512 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2513 | if (child->d_inode) |
2514 | return -EEXIST; | |
2515 | if (IS_DEADDIR(dir)) | |
2516 | return -ENOENT; | |
f419a2e3 | 2517 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2518 | } |
2519 | ||
1da177e4 LT |
2520 | /* |
2521 | * p1 and p2 should be directories on the same fs. | |
2522 | */ | |
2523 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2524 | { | |
2525 | struct dentry *p; | |
2526 | ||
2527 | if (p1 == p2) { | |
f2eace23 | 2528 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2529 | return NULL; |
2530 | } | |
2531 | ||
a11f3a05 | 2532 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2533 | |
e2761a11 OH |
2534 | p = d_ancestor(p2, p1); |
2535 | if (p) { | |
2536 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2537 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2538 | return p; | |
1da177e4 LT |
2539 | } |
2540 | ||
e2761a11 OH |
2541 | p = d_ancestor(p1, p2); |
2542 | if (p) { | |
2543 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2544 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2545 | return p; | |
1da177e4 LT |
2546 | } |
2547 | ||
f2eace23 | 2548 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
d1b72cc6 | 2549 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT2); |
1da177e4 LT |
2550 | return NULL; |
2551 | } | |
4d359507 | 2552 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2553 | |
2554 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2555 | { | |
1b1dcc1b | 2556 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2557 | if (p1 != p2) { |
1b1dcc1b | 2558 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2559 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2560 | } |
2561 | } | |
4d359507 | 2562 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2563 | |
4acdaf27 | 2564 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2565 | bool want_excl) |
1da177e4 | 2566 | { |
a95164d9 | 2567 | int error = may_create(dir, dentry); |
1da177e4 LT |
2568 | if (error) |
2569 | return error; | |
2570 | ||
acfa4380 | 2571 | if (!dir->i_op->create) |
1da177e4 LT |
2572 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2573 | mode &= S_IALLUGO; | |
2574 | mode |= S_IFREG; | |
2575 | error = security_inode_create(dir, dentry, mode); | |
2576 | if (error) | |
2577 | return error; | |
312b63fb | 2578 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2579 | if (!error) |
f38aa942 | 2580 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2581 | return error; |
2582 | } | |
4d359507 | 2583 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2584 | |
73d049a4 | 2585 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2586 | { |
3fb64190 | 2587 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2588 | struct inode *inode = dentry->d_inode; |
2589 | int error; | |
2590 | ||
bcda7652 AV |
2591 | /* O_PATH? */ |
2592 | if (!acc_mode) | |
2593 | return 0; | |
2594 | ||
1da177e4 LT |
2595 | if (!inode) |
2596 | return -ENOENT; | |
2597 | ||
c8fe8f30 CH |
2598 | switch (inode->i_mode & S_IFMT) { |
2599 | case S_IFLNK: | |
1da177e4 | 2600 | return -ELOOP; |
c8fe8f30 CH |
2601 | case S_IFDIR: |
2602 | if (acc_mode & MAY_WRITE) | |
2603 | return -EISDIR; | |
2604 | break; | |
2605 | case S_IFBLK: | |
2606 | case S_IFCHR: | |
3fb64190 | 2607 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2608 | return -EACCES; |
c8fe8f30 CH |
2609 | /*FALLTHRU*/ |
2610 | case S_IFIFO: | |
2611 | case S_IFSOCK: | |
1da177e4 | 2612 | flag &= ~O_TRUNC; |
c8fe8f30 | 2613 | break; |
4a3fd211 | 2614 | } |
b41572e9 | 2615 | |
3fb64190 | 2616 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2617 | if (error) |
2618 | return error; | |
6146f0d5 | 2619 | |
1da177e4 LT |
2620 | /* |
2621 | * An append-only file must be opened in append mode for writing. | |
2622 | */ | |
2623 | if (IS_APPEND(inode)) { | |
8737c930 | 2624 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2625 | return -EPERM; |
1da177e4 | 2626 | if (flag & O_TRUNC) |
7715b521 | 2627 | return -EPERM; |
1da177e4 LT |
2628 | } |
2629 | ||
2630 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2631 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2632 | return -EPERM; |
1da177e4 | 2633 | |
f3c7691e | 2634 | return 0; |
7715b521 | 2635 | } |
1da177e4 | 2636 | |
e1181ee6 | 2637 | static int handle_truncate(struct file *filp) |
7715b521 | 2638 | { |
e1181ee6 | 2639 | struct path *path = &filp->f_path; |
7715b521 AV |
2640 | struct inode *inode = path->dentry->d_inode; |
2641 | int error = get_write_access(inode); | |
2642 | if (error) | |
2643 | return error; | |
2644 | /* | |
2645 | * Refuse to truncate files with mandatory locks held on them. | |
2646 | */ | |
d7a06983 | 2647 | error = locks_verify_locked(filp); |
7715b521 | 2648 | if (!error) |
ea0d3ab2 | 2649 | error = security_path_truncate(path); |
7715b521 AV |
2650 | if (!error) { |
2651 | error = do_truncate(path->dentry, 0, | |
2652 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2653 | filp); |
7715b521 AV |
2654 | } |
2655 | put_write_access(inode); | |
acd0c935 | 2656 | return error; |
1da177e4 LT |
2657 | } |
2658 | ||
d57999e1 DH |
2659 | static inline int open_to_namei_flags(int flag) |
2660 | { | |
8a5e929d AV |
2661 | if ((flag & O_ACCMODE) == 3) |
2662 | flag--; | |
d57999e1 DH |
2663 | return flag; |
2664 | } | |
2665 | ||
d18e9008 MS |
2666 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2667 | { | |
2668 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2669 | if (error) | |
2670 | return error; | |
2671 | ||
2672 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2673 | if (error) | |
2674 | return error; | |
2675 | ||
2676 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2677 | } | |
2678 | ||
1acf0af9 DH |
2679 | /* |
2680 | * Attempt to atomically look up, create and open a file from a negative | |
2681 | * dentry. | |
2682 | * | |
2683 | * Returns 0 if successful. The file will have been created and attached to | |
2684 | * @file by the filesystem calling finish_open(). | |
2685 | * | |
2686 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2687 | * caller will need to perform the open themselves. @path will have been | |
2688 | * updated to point to the new dentry. This may be negative. | |
2689 | * | |
2690 | * Returns an error code otherwise. | |
2691 | */ | |
2675a4eb AV |
2692 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2693 | struct path *path, struct file *file, | |
2694 | const struct open_flags *op, | |
64894cf8 | 2695 | bool got_write, bool need_lookup, |
2675a4eb | 2696 | int *opened) |
d18e9008 MS |
2697 | { |
2698 | struct inode *dir = nd->path.dentry->d_inode; | |
2699 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2700 | umode_t mode; | |
2701 | int error; | |
2702 | int acc_mode; | |
d18e9008 MS |
2703 | int create_error = 0; |
2704 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
116cc022 | 2705 | bool excl; |
d18e9008 MS |
2706 | |
2707 | BUG_ON(dentry->d_inode); | |
2708 | ||
2709 | /* Don't create child dentry for a dead directory. */ | |
2710 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2711 | error = -ENOENT; |
d18e9008 MS |
2712 | goto out; |
2713 | } | |
2714 | ||
62b259d8 | 2715 | mode = op->mode; |
d18e9008 MS |
2716 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2717 | mode &= ~current_umask(); | |
2718 | ||
116cc022 MS |
2719 | excl = (open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT); |
2720 | if (excl) | |
d18e9008 | 2721 | open_flag &= ~O_TRUNC; |
d18e9008 MS |
2722 | |
2723 | /* | |
2724 | * Checking write permission is tricky, bacuse we don't know if we are | |
2725 | * going to actually need it: O_CREAT opens should work as long as the | |
2726 | * file exists. But checking existence breaks atomicity. The trick is | |
2727 | * to check access and if not granted clear O_CREAT from the flags. | |
2728 | * | |
2729 | * Another problem is returing the "right" error value (e.g. for an | |
2730 | * O_EXCL open we want to return EEXIST not EROFS). | |
2731 | */ | |
64894cf8 AV |
2732 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2733 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2734 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2735 | /* |
2736 | * No O_CREATE -> atomicity not a requirement -> fall | |
2737 | * back to lookup + open | |
2738 | */ | |
2739 | goto no_open; | |
2740 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2741 | /* Fall back and fail with the right error */ | |
64894cf8 | 2742 | create_error = -EROFS; |
d18e9008 MS |
2743 | goto no_open; |
2744 | } else { | |
2745 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2746 | create_error = -EROFS; |
d18e9008 MS |
2747 | open_flag &= ~O_CREAT; |
2748 | } | |
2749 | } | |
2750 | ||
2751 | if (open_flag & O_CREAT) { | |
38227f78 | 2752 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2753 | if (error) { |
2754 | create_error = error; | |
2755 | if (open_flag & O_EXCL) | |
2756 | goto no_open; | |
2757 | open_flag &= ~O_CREAT; | |
2758 | } | |
2759 | } | |
2760 | ||
2761 | if (nd->flags & LOOKUP_DIRECTORY) | |
2762 | open_flag |= O_DIRECTORY; | |
2763 | ||
30d90494 AV |
2764 | file->f_path.dentry = DENTRY_NOT_SET; |
2765 | file->f_path.mnt = nd->path.mnt; | |
2766 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2767 | opened); |
d9585277 | 2768 | if (error < 0) { |
d9585277 AV |
2769 | if (create_error && error == -ENOENT) |
2770 | error = create_error; | |
d18e9008 MS |
2771 | goto out; |
2772 | } | |
2773 | ||
d9585277 | 2774 | if (error) { /* returned 1, that is */ |
30d90494 | 2775 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2776 | error = -EIO; |
d18e9008 MS |
2777 | goto out; |
2778 | } | |
30d90494 | 2779 | if (file->f_path.dentry) { |
d18e9008 | 2780 | dput(dentry); |
30d90494 | 2781 | dentry = file->f_path.dentry; |
d18e9008 | 2782 | } |
03da633a AV |
2783 | if (*opened & FILE_CREATED) |
2784 | fsnotify_create(dir, dentry); | |
2785 | if (!dentry->d_inode) { | |
2786 | WARN_ON(*opened & FILE_CREATED); | |
2787 | if (create_error) { | |
2788 | error = create_error; | |
2789 | goto out; | |
2790 | } | |
2791 | } else { | |
2792 | if (excl && !(*opened & FILE_CREATED)) { | |
2793 | error = -EEXIST; | |
2794 | goto out; | |
2795 | } | |
62b2ce96 | 2796 | } |
d18e9008 MS |
2797 | goto looked_up; |
2798 | } | |
2799 | ||
2800 | /* | |
2801 | * We didn't have the inode before the open, so check open permission | |
2802 | * here. | |
2803 | */ | |
03da633a AV |
2804 | acc_mode = op->acc_mode; |
2805 | if (*opened & FILE_CREATED) { | |
2806 | WARN_ON(!(open_flag & O_CREAT)); | |
2807 | fsnotify_create(dir, dentry); | |
2808 | acc_mode = MAY_OPEN; | |
2809 | } | |
2675a4eb AV |
2810 | error = may_open(&file->f_path, acc_mode, open_flag); |
2811 | if (error) | |
2812 | fput(file); | |
d18e9008 MS |
2813 | |
2814 | out: | |
2815 | dput(dentry); | |
2675a4eb | 2816 | return error; |
d18e9008 | 2817 | |
d18e9008 MS |
2818 | no_open: |
2819 | if (need_lookup) { | |
72bd866a | 2820 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2821 | if (IS_ERR(dentry)) |
2675a4eb | 2822 | return PTR_ERR(dentry); |
d18e9008 MS |
2823 | |
2824 | if (create_error) { | |
2825 | int open_flag = op->open_flag; | |
2826 | ||
2675a4eb | 2827 | error = create_error; |
d18e9008 MS |
2828 | if ((open_flag & O_EXCL)) { |
2829 | if (!dentry->d_inode) | |
2830 | goto out; | |
2831 | } else if (!dentry->d_inode) { | |
2832 | goto out; | |
2833 | } else if ((open_flag & O_TRUNC) && | |
e36cb0b8 | 2834 | d_is_reg(dentry)) { |
d18e9008 MS |
2835 | goto out; |
2836 | } | |
2837 | /* will fail later, go on to get the right error */ | |
2838 | } | |
2839 | } | |
2840 | looked_up: | |
2841 | path->dentry = dentry; | |
2842 | path->mnt = nd->path.mnt; | |
2675a4eb | 2843 | return 1; |
d18e9008 MS |
2844 | } |
2845 | ||
d58ffd35 | 2846 | /* |
1acf0af9 | 2847 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2848 | * |
2849 | * Must be called with i_mutex held on parent. | |
2850 | * | |
1acf0af9 DH |
2851 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2852 | * opened. In this case the file will be returned attached to @file. | |
2853 | * | |
2854 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2855 | * and creations will have been performed and the dentry returned in @path will | |
2856 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2857 | * specified then a negative dentry may be returned. | |
2858 | * | |
2859 | * An error code is returned otherwise. | |
2860 | * | |
2861 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2862 | * cleared otherwise prior to returning. | |
d58ffd35 | 2863 | */ |
2675a4eb AV |
2864 | static int lookup_open(struct nameidata *nd, struct path *path, |
2865 | struct file *file, | |
2866 | const struct open_flags *op, | |
64894cf8 | 2867 | bool got_write, int *opened) |
d58ffd35 MS |
2868 | { |
2869 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2870 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2871 | struct dentry *dentry; |
2872 | int error; | |
54ef4872 | 2873 | bool need_lookup; |
d58ffd35 | 2874 | |
47237687 | 2875 | *opened &= ~FILE_CREATED; |
201f956e | 2876 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2877 | if (IS_ERR(dentry)) |
2675a4eb | 2878 | return PTR_ERR(dentry); |
d58ffd35 | 2879 | |
d18e9008 MS |
2880 | /* Cached positive dentry: will open in f_op->open */ |
2881 | if (!need_lookup && dentry->d_inode) | |
2882 | goto out_no_open; | |
2883 | ||
2884 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 2885 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 2886 | need_lookup, opened); |
d18e9008 MS |
2887 | } |
2888 | ||
54ef4872 MS |
2889 | if (need_lookup) { |
2890 | BUG_ON(dentry->d_inode); | |
2891 | ||
72bd866a | 2892 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2893 | if (IS_ERR(dentry)) |
2675a4eb | 2894 | return PTR_ERR(dentry); |
54ef4872 MS |
2895 | } |
2896 | ||
d58ffd35 MS |
2897 | /* Negative dentry, just create the file */ |
2898 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2899 | umode_t mode = op->mode; | |
2900 | if (!IS_POSIXACL(dir->d_inode)) | |
2901 | mode &= ~current_umask(); | |
2902 | /* | |
2903 | * This write is needed to ensure that a | |
2904 | * rw->ro transition does not occur between | |
2905 | * the time when the file is created and when | |
2906 | * a permanent write count is taken through | |
015c3bbc | 2907 | * the 'struct file' in finish_open(). |
d58ffd35 | 2908 | */ |
64894cf8 AV |
2909 | if (!got_write) { |
2910 | error = -EROFS; | |
d58ffd35 | 2911 | goto out_dput; |
64894cf8 | 2912 | } |
47237687 | 2913 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2914 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2915 | if (error) | |
2916 | goto out_dput; | |
312b63fb AV |
2917 | error = vfs_create(dir->d_inode, dentry, mode, |
2918 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2919 | if (error) |
2920 | goto out_dput; | |
2921 | } | |
d18e9008 | 2922 | out_no_open: |
d58ffd35 MS |
2923 | path->dentry = dentry; |
2924 | path->mnt = nd->path.mnt; | |
2675a4eb | 2925 | return 1; |
d58ffd35 MS |
2926 | |
2927 | out_dput: | |
2928 | dput(dentry); | |
2675a4eb | 2929 | return error; |
d58ffd35 MS |
2930 | } |
2931 | ||
31e6b01f | 2932 | /* |
fe2d35ff | 2933 | * Handle the last step of open() |
31e6b01f | 2934 | */ |
896475d5 | 2935 | static int do_last(struct nameidata *nd, |
2675a4eb | 2936 | struct file *file, const struct open_flags *op, |
669abf4e | 2937 | int *opened, struct filename *name) |
fb1cc555 | 2938 | { |
a1e28038 | 2939 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 2940 | int open_flag = op->open_flag; |
77d660a8 | 2941 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 2942 | bool got_write = false; |
bcda7652 | 2943 | int acc_mode = op->acc_mode; |
a1eb3315 | 2944 | struct inode *inode; |
16b1c1cd | 2945 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
896475d5 | 2946 | struct path path; |
16b1c1cd | 2947 | bool retried = false; |
16c2cd71 | 2948 | int error; |
1f36f774 | 2949 | |
c3e380b0 AV |
2950 | nd->flags &= ~LOOKUP_PARENT; |
2951 | nd->flags |= op->intent; | |
2952 | ||
bc77daa7 | 2953 | if (nd->last_type != LAST_NORM) { |
fe2d35ff AV |
2954 | error = handle_dots(nd, nd->last_type); |
2955 | if (error) | |
2675a4eb | 2956 | return error; |
e83db167 | 2957 | goto finish_open; |
1f36f774 | 2958 | } |
67ee3ad2 | 2959 | |
ca344a89 | 2960 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
2961 | if (nd->last.name[nd->last.len]) |
2962 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2963 | /* we _can_ be in RCU mode here */ | |
896475d5 | 2964 | error = lookup_fast(nd, &path, &inode); |
71574865 MS |
2965 | if (likely(!error)) |
2966 | goto finish_lookup; | |
2967 | ||
2968 | if (error < 0) | |
2675a4eb | 2969 | goto out; |
71574865 MS |
2970 | |
2971 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
2972 | } else { |
2973 | /* create side of things */ | |
2974 | /* | |
2975 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
2976 | * has been cleared when we got to the last component we are | |
2977 | * about to look up | |
2978 | */ | |
2979 | error = complete_walk(nd); | |
2980 | if (error) | |
2675a4eb | 2981 | return error; |
fe2d35ff | 2982 | |
33e2208a | 2983 | audit_inode(name, dir, LOOKUP_PARENT); |
b6183df7 MS |
2984 | error = -EISDIR; |
2985 | /* trailing slashes? */ | |
2986 | if (nd->last.name[nd->last.len]) | |
2675a4eb | 2987 | goto out; |
b6183df7 | 2988 | } |
a2c36b45 | 2989 | |
16b1c1cd | 2990 | retry_lookup: |
64894cf8 AV |
2991 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
2992 | error = mnt_want_write(nd->path.mnt); | |
2993 | if (!error) | |
2994 | got_write = true; | |
2995 | /* | |
2996 | * do _not_ fail yet - we might not need that or fail with | |
2997 | * a different error; let lookup_open() decide; we'll be | |
2998 | * dropping this one anyway. | |
2999 | */ | |
3000 | } | |
a1e28038 | 3001 | mutex_lock(&dir->d_inode->i_mutex); |
896475d5 | 3002 | error = lookup_open(nd, &path, file, op, got_write, opened); |
d58ffd35 | 3003 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 3004 | |
2675a4eb AV |
3005 | if (error <= 0) { |
3006 | if (error) | |
d18e9008 MS |
3007 | goto out; |
3008 | ||
47237687 | 3009 | if ((*opened & FILE_CREATED) || |
496ad9aa | 3010 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3011 | will_truncate = false; |
d18e9008 | 3012 | |
adb5c247 | 3013 | audit_inode(name, file->f_path.dentry, 0); |
d18e9008 MS |
3014 | goto opened; |
3015 | } | |
fb1cc555 | 3016 | |
47237687 | 3017 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 3018 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3019 | open_flag &= ~O_TRUNC; |
77d660a8 | 3020 | will_truncate = false; |
bcda7652 | 3021 | acc_mode = MAY_OPEN; |
896475d5 | 3022 | path_to_nameidata(&path, nd); |
e83db167 | 3023 | goto finish_open_created; |
fb1cc555 AV |
3024 | } |
3025 | ||
3026 | /* | |
3134f37e | 3027 | * create/update audit record if it already exists. |
fb1cc555 | 3028 | */ |
896475d5 AV |
3029 | if (d_is_positive(path.dentry)) |
3030 | audit_inode(name, path.dentry, 0); | |
fb1cc555 | 3031 | |
d18e9008 MS |
3032 | /* |
3033 | * If atomic_open() acquired write access it is dropped now due to | |
3034 | * possible mount and symlink following (this might be optimized away if | |
3035 | * necessary...) | |
3036 | */ | |
64894cf8 | 3037 | if (got_write) { |
d18e9008 | 3038 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3039 | got_write = false; |
d18e9008 MS |
3040 | } |
3041 | ||
fb1cc555 | 3042 | error = -EEXIST; |
f8310c59 | 3043 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) |
fb1cc555 AV |
3044 | goto exit_dput; |
3045 | ||
896475d5 | 3046 | error = follow_managed(&path, nd->flags); |
9875cf80 DH |
3047 | if (error < 0) |
3048 | goto exit_dput; | |
fb1cc555 | 3049 | |
a3fbbde7 AV |
3050 | if (error) |
3051 | nd->flags |= LOOKUP_JUMPED; | |
3052 | ||
decf3400 | 3053 | BUG_ON(nd->flags & LOOKUP_RCU); |
896475d5 | 3054 | inode = path.dentry->d_inode; |
fb1cc555 | 3055 | error = -ENOENT; |
896475d5 AV |
3056 | if (d_is_negative(path.dentry)) { |
3057 | path_to_nameidata(&path, nd); | |
2675a4eb | 3058 | goto out; |
54c33e7f | 3059 | } |
766c4cbf | 3060 | finish_lookup: |
896475d5 | 3061 | if (should_follow_link(path.dentry, nd->flags & LOOKUP_FOLLOW)) { |
d45ea867 | 3062 | if (nd->flags & LOOKUP_RCU) { |
896475d5 AV |
3063 | if (unlikely(nd->path.mnt != path.mnt || |
3064 | unlazy_walk(nd, path.dentry))) { | |
d45ea867 | 3065 | error = -ECHILD; |
2675a4eb | 3066 | goto out; |
d45ea867 MS |
3067 | } |
3068 | } | |
896475d5 AV |
3069 | BUG_ON(inode != path.dentry->d_inode); |
3070 | nd->link = path; | |
2675a4eb | 3071 | return 1; |
d45ea867 | 3072 | } |
fb1cc555 | 3073 | |
896475d5 AV |
3074 | if (unlikely(d_is_symlink(path.dentry)) && !(open_flag & O_PATH)) { |
3075 | path_to_nameidata(&path, nd); | |
a5cfe2d5 AV |
3076 | error = -ELOOP; |
3077 | goto out; | |
3078 | } | |
3079 | ||
896475d5 AV |
3080 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path.mnt) { |
3081 | path_to_nameidata(&path, nd); | |
16b1c1cd MS |
3082 | } else { |
3083 | save_parent.dentry = nd->path.dentry; | |
896475d5 AV |
3084 | save_parent.mnt = mntget(path.mnt); |
3085 | nd->path.dentry = path.dentry; | |
16b1c1cd MS |
3086 | |
3087 | } | |
decf3400 | 3088 | nd->inode = inode; |
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); |
e276ae67 | 3143 | terminate_walk(nd); |
2675a4eb | 3144 | return error; |
fb1cc555 | 3145 | |
fb1cc555 | 3146 | exit_dput: |
896475d5 | 3147 | path_put_conditional(&path, nd); |
ca344a89 | 3148 | goto out; |
015c3bbc | 3149 | exit_fput: |
2675a4eb AV |
3150 | fput(file); |
3151 | goto out; | |
015c3bbc | 3152 | |
f60dc3db MS |
3153 | stale_open: |
3154 | /* If no saved parent or already retried then can't retry */ | |
3155 | if (!save_parent.dentry || retried) | |
3156 | goto out; | |
3157 | ||
3158 | BUG_ON(save_parent.dentry != dir); | |
3159 | path_put(&nd->path); | |
3160 | nd->path = save_parent; | |
3161 | nd->inode = dir->d_inode; | |
3162 | save_parent.mnt = NULL; | |
3163 | save_parent.dentry = NULL; | |
64894cf8 | 3164 | if (got_write) { |
f60dc3db | 3165 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3166 | got_write = false; |
f60dc3db MS |
3167 | } |
3168 | retried = true; | |
3169 | goto retry_lookup; | |
fb1cc555 AV |
3170 | } |
3171 | ||
60545d0d AV |
3172 | static int do_tmpfile(int dfd, struct filename *pathname, |
3173 | struct nameidata *nd, int flags, | |
3174 | const struct open_flags *op, | |
3175 | struct file *file, int *opened) | |
3176 | { | |
3177 | static const struct qstr name = QSTR_INIT("/", 1); | |
3178 | struct dentry *dentry, *child; | |
3179 | struct inode *dir; | |
5eb6b495 | 3180 | int error = path_lookupat(dfd, pathname, |
60545d0d AV |
3181 | flags | LOOKUP_DIRECTORY, nd); |
3182 | if (unlikely(error)) | |
3183 | return error; | |
3184 | error = mnt_want_write(nd->path.mnt); | |
3185 | if (unlikely(error)) | |
3186 | goto out; | |
3187 | /* we want directory to be writable */ | |
3188 | error = inode_permission(nd->inode, MAY_WRITE | MAY_EXEC); | |
3189 | if (error) | |
3190 | goto out2; | |
3191 | dentry = nd->path.dentry; | |
3192 | dir = dentry->d_inode; | |
3193 | if (!dir->i_op->tmpfile) { | |
3194 | error = -EOPNOTSUPP; | |
3195 | goto out2; | |
3196 | } | |
3197 | child = d_alloc(dentry, &name); | |
3198 | if (unlikely(!child)) { | |
3199 | error = -ENOMEM; | |
3200 | goto out2; | |
3201 | } | |
3202 | nd->flags &= ~LOOKUP_DIRECTORY; | |
3203 | nd->flags |= op->intent; | |
3204 | dput(nd->path.dentry); | |
3205 | nd->path.dentry = child; | |
3206 | error = dir->i_op->tmpfile(dir, nd->path.dentry, op->mode); | |
3207 | if (error) | |
3208 | goto out2; | |
3209 | audit_inode(pathname, nd->path.dentry, 0); | |
69a91c23 ER |
3210 | /* Don't check for other permissions, the inode was just created */ |
3211 | error = may_open(&nd->path, MAY_OPEN, op->open_flag); | |
60545d0d AV |
3212 | if (error) |
3213 | goto out2; | |
3214 | file->f_path.mnt = nd->path.mnt; | |
3215 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
3216 | if (error) | |
3217 | goto out2; | |
3218 | error = open_check_o_direct(file); | |
f4e0c30c | 3219 | if (error) { |
60545d0d | 3220 | fput(file); |
f4e0c30c AV |
3221 | } else if (!(op->open_flag & O_EXCL)) { |
3222 | struct inode *inode = file_inode(file); | |
3223 | spin_lock(&inode->i_lock); | |
3224 | inode->i_state |= I_LINKABLE; | |
3225 | spin_unlock(&inode->i_lock); | |
3226 | } | |
60545d0d AV |
3227 | out2: |
3228 | mnt_drop_write(nd->path.mnt); | |
3229 | out: | |
3230 | path_put(&nd->path); | |
3231 | return error; | |
3232 | } | |
3233 | ||
669abf4e | 3234 | static struct file *path_openat(int dfd, struct filename *pathname, |
73d049a4 | 3235 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 3236 | { |
30d90494 | 3237 | struct file *file; |
47237687 | 3238 | int opened = 0; |
13aab428 | 3239 | int error; |
31e6b01f | 3240 | |
30d90494 | 3241 | file = get_empty_filp(); |
1afc99be AV |
3242 | if (IS_ERR(file)) |
3243 | return file; | |
31e6b01f | 3244 | |
30d90494 | 3245 | file->f_flags = op->open_flag; |
31e6b01f | 3246 | |
bb458c64 | 3247 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
60545d0d | 3248 | error = do_tmpfile(dfd, pathname, nd, flags, op, file, &opened); |
f15133df | 3249 | goto out2; |
60545d0d AV |
3250 | } |
3251 | ||
6e8a1f87 | 3252 | error = path_init(dfd, pathname, flags, nd); |
31e6b01f | 3253 | if (unlikely(error)) |
2675a4eb | 3254 | goto out; |
1da177e4 | 3255 | |
896475d5 | 3256 | error = do_last(nd, file, op, &opened, pathname); |
2675a4eb | 3257 | while (unlikely(error > 0)) { /* trailing symlink */ |
73d049a4 | 3258 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); |
1d8e03d3 | 3259 | error = trailing_symlink(nd); |
c3e380b0 | 3260 | if (unlikely(error)) |
2675a4eb | 3261 | break; |
896475d5 | 3262 | error = do_last(nd, file, op, &opened, pathname); |
b9ff4429 | 3263 | put_link(nd); |
806b681c | 3264 | } |
10fa8e62 | 3265 | out: |
893b7775 | 3266 | path_cleanup(nd); |
f15133df | 3267 | out2: |
2675a4eb AV |
3268 | if (!(opened & FILE_OPENED)) { |
3269 | BUG_ON(!error); | |
30d90494 | 3270 | put_filp(file); |
16b1c1cd | 3271 | } |
2675a4eb AV |
3272 | if (unlikely(error)) { |
3273 | if (error == -EOPENSTALE) { | |
3274 | if (flags & LOOKUP_RCU) | |
3275 | error = -ECHILD; | |
3276 | else | |
3277 | error = -ESTALE; | |
3278 | } | |
3279 | file = ERR_PTR(error); | |
3280 | } | |
3281 | return file; | |
1da177e4 LT |
3282 | } |
3283 | ||
669abf4e | 3284 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3285 | const struct open_flags *op) |
13aab428 | 3286 | { |
73d049a4 | 3287 | struct nameidata nd; |
f9652e10 | 3288 | int flags = op->lookup_flags; |
13aab428 AV |
3289 | struct file *filp; |
3290 | ||
73d049a4 | 3291 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 3292 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 3293 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 3294 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 3295 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
3296 | return filp; |
3297 | } | |
3298 | ||
73d049a4 | 3299 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3300 | const char *name, const struct open_flags *op) |
73d049a4 AV |
3301 | { |
3302 | struct nameidata nd; | |
3303 | struct file *file; | |
51689104 | 3304 | struct filename *filename; |
f9652e10 | 3305 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3306 | |
3307 | nd.root.mnt = mnt; | |
3308 | nd.root.dentry = dentry; | |
3309 | ||
b18825a7 | 3310 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3311 | return ERR_PTR(-ELOOP); |
3312 | ||
51689104 PM |
3313 | filename = getname_kernel(name); |
3314 | if (unlikely(IS_ERR(filename))) | |
3315 | return ERR_CAST(filename); | |
3316 | ||
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 PM |
3321 | file = path_openat(-1, filename, &nd, op, flags | LOOKUP_REVAL); |
3322 | putname(filename); | |
73d049a4 AV |
3323 | return file; |
3324 | } | |
3325 | ||
fa14a0b8 | 3326 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3327 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3328 | { |
c663e5d8 | 3329 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d | 3330 | struct nameidata nd; |
c30dabfe | 3331 | int err2; |
1ac12b4b JL |
3332 | int error; |
3333 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3334 | ||
3335 | /* | |
3336 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3337 | * other flags passed in are ignored! | |
3338 | */ | |
3339 | lookup_flags &= LOOKUP_REVAL; | |
3340 | ||
fa14a0b8 | 3341 | error = filename_lookup(dfd, name, LOOKUP_PARENT|lookup_flags, &nd); |
ed75e95d AV |
3342 | if (error) |
3343 | return ERR_PTR(error); | |
1da177e4 | 3344 | |
c663e5d8 CH |
3345 | /* |
3346 | * Yucky last component or no last component at all? | |
3347 | * (foo/., foo/.., /////) | |
3348 | */ | |
ed75e95d AV |
3349 | if (nd.last_type != LAST_NORM) |
3350 | goto out; | |
3351 | nd.flags &= ~LOOKUP_PARENT; | |
3352 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 | 3353 | |
c30dabfe JK |
3354 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
3355 | err2 = mnt_want_write(nd.path.mnt); | |
c663e5d8 CH |
3356 | /* |
3357 | * Do the final lookup. | |
3358 | */ | |
ed75e95d | 3359 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
6a9f40d6 | 3360 | dentry = __lookup_hash(&nd.last, nd.path.dentry, nd.flags); |
1da177e4 | 3361 | if (IS_ERR(dentry)) |
a8104a9f | 3362 | goto unlock; |
c663e5d8 | 3363 | |
a8104a9f | 3364 | error = -EEXIST; |
b18825a7 | 3365 | if (d_is_positive(dentry)) |
a8104a9f | 3366 | goto fail; |
b18825a7 | 3367 | |
c663e5d8 CH |
3368 | /* |
3369 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3370 | * From the vfs_mknod() POV we just have a negative dentry - | |
3371 | * all is fine. Let's be bastards - you had / on the end, you've | |
3372 | * been asking for (non-existent) directory. -ENOENT for you. | |
3373 | */ | |
ed75e95d | 3374 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
a8104a9f | 3375 | error = -ENOENT; |
ed75e95d | 3376 | goto fail; |
e9baf6e5 | 3377 | } |
c30dabfe JK |
3378 | if (unlikely(err2)) { |
3379 | error = err2; | |
a8104a9f | 3380 | goto fail; |
c30dabfe | 3381 | } |
ed75e95d | 3382 | *path = nd.path; |
1da177e4 | 3383 | return dentry; |
1da177e4 | 3384 | fail: |
a8104a9f AV |
3385 | dput(dentry); |
3386 | dentry = ERR_PTR(error); | |
3387 | unlock: | |
ed75e95d | 3388 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe JK |
3389 | if (!err2) |
3390 | mnt_drop_write(nd.path.mnt); | |
ed75e95d AV |
3391 | out: |
3392 | path_put(&nd.path); | |
1da177e4 LT |
3393 | return dentry; |
3394 | } | |
fa14a0b8 AV |
3395 | |
3396 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3397 | struct path *path, unsigned int lookup_flags) | |
3398 | { | |
51689104 PM |
3399 | struct filename *filename = getname_kernel(pathname); |
3400 | struct dentry *res; | |
3401 | ||
3402 | if (IS_ERR(filename)) | |
3403 | return ERR_CAST(filename); | |
3404 | res = filename_create(dfd, filename, path, lookup_flags); | |
3405 | putname(filename); | |
3406 | return res; | |
fa14a0b8 | 3407 | } |
dae6ad8f AV |
3408 | EXPORT_SYMBOL(kern_path_create); |
3409 | ||
921a1650 AV |
3410 | void done_path_create(struct path *path, struct dentry *dentry) |
3411 | { | |
3412 | dput(dentry); | |
3413 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
a8104a9f | 3414 | mnt_drop_write(path->mnt); |
921a1650 AV |
3415 | path_put(path); |
3416 | } | |
3417 | EXPORT_SYMBOL(done_path_create); | |
3418 | ||
1ac12b4b JL |
3419 | struct dentry *user_path_create(int dfd, const char __user *pathname, |
3420 | struct path *path, unsigned int lookup_flags) | |
dae6ad8f | 3421 | { |
91a27b2a | 3422 | struct filename *tmp = getname(pathname); |
dae6ad8f AV |
3423 | struct dentry *res; |
3424 | if (IS_ERR(tmp)) | |
3425 | return ERR_CAST(tmp); | |
fa14a0b8 | 3426 | res = filename_create(dfd, tmp, path, lookup_flags); |
dae6ad8f AV |
3427 | putname(tmp); |
3428 | return res; | |
3429 | } | |
3430 | EXPORT_SYMBOL(user_path_create); | |
3431 | ||
1a67aafb | 3432 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3433 | { |
a95164d9 | 3434 | int error = may_create(dir, dentry); |
1da177e4 LT |
3435 | |
3436 | if (error) | |
3437 | return error; | |
3438 | ||
975d6b39 | 3439 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3440 | return -EPERM; |
3441 | ||
acfa4380 | 3442 | if (!dir->i_op->mknod) |
1da177e4 LT |
3443 | return -EPERM; |
3444 | ||
08ce5f16 SH |
3445 | error = devcgroup_inode_mknod(mode, dev); |
3446 | if (error) | |
3447 | return error; | |
3448 | ||
1da177e4 LT |
3449 | error = security_inode_mknod(dir, dentry, mode, dev); |
3450 | if (error) | |
3451 | return error; | |
3452 | ||
1da177e4 | 3453 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3454 | if (!error) |
f38aa942 | 3455 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3456 | return error; |
3457 | } | |
4d359507 | 3458 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3459 | |
f69aac00 | 3460 | static int may_mknod(umode_t mode) |
463c3197 DH |
3461 | { |
3462 | switch (mode & S_IFMT) { | |
3463 | case S_IFREG: | |
3464 | case S_IFCHR: | |
3465 | case S_IFBLK: | |
3466 | case S_IFIFO: | |
3467 | case S_IFSOCK: | |
3468 | case 0: /* zero mode translates to S_IFREG */ | |
3469 | return 0; | |
3470 | case S_IFDIR: | |
3471 | return -EPERM; | |
3472 | default: | |
3473 | return -EINVAL; | |
3474 | } | |
3475 | } | |
3476 | ||
8208a22b | 3477 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3478 | unsigned, dev) |
1da177e4 | 3479 | { |
2ad94ae6 | 3480 | struct dentry *dentry; |
dae6ad8f AV |
3481 | struct path path; |
3482 | int error; | |
972567f1 | 3483 | unsigned int lookup_flags = 0; |
1da177e4 | 3484 | |
8e4bfca1 AV |
3485 | error = may_mknod(mode); |
3486 | if (error) | |
3487 | return error; | |
972567f1 JL |
3488 | retry: |
3489 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3490 | if (IS_ERR(dentry)) |
3491 | return PTR_ERR(dentry); | |
2ad94ae6 | 3492 | |
dae6ad8f | 3493 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3494 | mode &= ~current_umask(); |
dae6ad8f | 3495 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3496 | if (error) |
a8104a9f | 3497 | goto out; |
463c3197 | 3498 | switch (mode & S_IFMT) { |
1da177e4 | 3499 | case 0: case S_IFREG: |
312b63fb | 3500 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3501 | break; |
3502 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3503 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3504 | new_decode_dev(dev)); |
3505 | break; | |
3506 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3507 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3508 | break; |
1da177e4 | 3509 | } |
a8104a9f | 3510 | out: |
921a1650 | 3511 | done_path_create(&path, dentry); |
972567f1 JL |
3512 | if (retry_estale(error, lookup_flags)) { |
3513 | lookup_flags |= LOOKUP_REVAL; | |
3514 | goto retry; | |
3515 | } | |
1da177e4 LT |
3516 | return error; |
3517 | } | |
3518 | ||
8208a22b | 3519 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3520 | { |
3521 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3522 | } | |
3523 | ||
18bb1db3 | 3524 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3525 | { |
a95164d9 | 3526 | int error = may_create(dir, dentry); |
8de52778 | 3527 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3528 | |
3529 | if (error) | |
3530 | return error; | |
3531 | ||
acfa4380 | 3532 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3533 | return -EPERM; |
3534 | ||
3535 | mode &= (S_IRWXUGO|S_ISVTX); | |
3536 | error = security_inode_mkdir(dir, dentry, mode); | |
3537 | if (error) | |
3538 | return error; | |
3539 | ||
8de52778 AV |
3540 | if (max_links && dir->i_nlink >= max_links) |
3541 | return -EMLINK; | |
3542 | ||
1da177e4 | 3543 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3544 | if (!error) |
f38aa942 | 3545 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3546 | return error; |
3547 | } | |
4d359507 | 3548 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3549 | |
a218d0fd | 3550 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3551 | { |
6902d925 | 3552 | struct dentry *dentry; |
dae6ad8f AV |
3553 | struct path path; |
3554 | int error; | |
b76d8b82 | 3555 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3556 | |
b76d8b82 JL |
3557 | retry: |
3558 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3559 | if (IS_ERR(dentry)) |
dae6ad8f | 3560 | return PTR_ERR(dentry); |
1da177e4 | 3561 | |
dae6ad8f | 3562 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3563 | mode &= ~current_umask(); |
dae6ad8f | 3564 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3565 | if (!error) |
3566 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3567 | done_path_create(&path, dentry); |
b76d8b82 JL |
3568 | if (retry_estale(error, lookup_flags)) { |
3569 | lookup_flags |= LOOKUP_REVAL; | |
3570 | goto retry; | |
3571 | } | |
1da177e4 LT |
3572 | return error; |
3573 | } | |
3574 | ||
a218d0fd | 3575 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3576 | { |
3577 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3578 | } | |
3579 | ||
1da177e4 | 3580 | /* |
a71905f0 | 3581 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3582 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3583 | * dentry, and if that is true (possibly after pruning the dcache), |
3584 | * then we drop the dentry now. | |
1da177e4 LT |
3585 | * |
3586 | * A low-level filesystem can, if it choses, legally | |
3587 | * do a | |
3588 | * | |
3589 | * if (!d_unhashed(dentry)) | |
3590 | * return -EBUSY; | |
3591 | * | |
3592 | * if it cannot handle the case of removing a directory | |
3593 | * that is still in use by something else.. | |
3594 | */ | |
3595 | void dentry_unhash(struct dentry *dentry) | |
3596 | { | |
dc168427 | 3597 | shrink_dcache_parent(dentry); |
1da177e4 | 3598 | spin_lock(&dentry->d_lock); |
98474236 | 3599 | if (dentry->d_lockref.count == 1) |
1da177e4 LT |
3600 | __d_drop(dentry); |
3601 | spin_unlock(&dentry->d_lock); | |
1da177e4 | 3602 | } |
4d359507 | 3603 | EXPORT_SYMBOL(dentry_unhash); |
1da177e4 LT |
3604 | |
3605 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3606 | { | |
3607 | int error = may_delete(dir, dentry, 1); | |
3608 | ||
3609 | if (error) | |
3610 | return error; | |
3611 | ||
acfa4380 | 3612 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3613 | return -EPERM; |
3614 | ||
1d2ef590 | 3615 | dget(dentry); |
1b1dcc1b | 3616 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3617 | |
3618 | error = -EBUSY; | |
7af1364f | 3619 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3620 | goto out; |
3621 | ||
3622 | error = security_inode_rmdir(dir, dentry); | |
3623 | if (error) | |
3624 | goto out; | |
3625 | ||
3cebde24 | 3626 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3627 | error = dir->i_op->rmdir(dir, dentry); |
3628 | if (error) | |
3629 | goto out; | |
3630 | ||
3631 | dentry->d_inode->i_flags |= S_DEAD; | |
3632 | dont_mount(dentry); | |
8ed936b5 | 3633 | detach_mounts(dentry); |
912dbc15 SW |
3634 | |
3635 | out: | |
1b1dcc1b | 3636 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3637 | dput(dentry); |
912dbc15 | 3638 | if (!error) |
1da177e4 | 3639 | d_delete(dentry); |
1da177e4 LT |
3640 | return error; |
3641 | } | |
4d359507 | 3642 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3643 | |
5590ff0d | 3644 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3645 | { |
3646 | int error = 0; | |
91a27b2a | 3647 | struct filename *name; |
1da177e4 | 3648 | struct dentry *dentry; |
f5beed75 AV |
3649 | struct path path; |
3650 | struct qstr last; | |
3651 | int type; | |
c6ee9206 JL |
3652 | unsigned int lookup_flags = 0; |
3653 | retry: | |
f5beed75 AV |
3654 | name = user_path_parent(dfd, pathname, |
3655 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3656 | if (IS_ERR(name)) |
3657 | return PTR_ERR(name); | |
1da177e4 | 3658 | |
f5beed75 | 3659 | switch (type) { |
0612d9fb OH |
3660 | case LAST_DOTDOT: |
3661 | error = -ENOTEMPTY; | |
3662 | goto exit1; | |
3663 | case LAST_DOT: | |
3664 | error = -EINVAL; | |
3665 | goto exit1; | |
3666 | case LAST_ROOT: | |
3667 | error = -EBUSY; | |
3668 | goto exit1; | |
1da177e4 | 3669 | } |
0612d9fb | 3670 | |
f5beed75 | 3671 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3672 | if (error) |
3673 | goto exit1; | |
0612d9fb | 3674 | |
f5beed75 AV |
3675 | mutex_lock_nested(&path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3676 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); | |
1da177e4 | 3677 | error = PTR_ERR(dentry); |
6902d925 DH |
3678 | if (IS_ERR(dentry)) |
3679 | goto exit2; | |
e6bc45d6 TT |
3680 | if (!dentry->d_inode) { |
3681 | error = -ENOENT; | |
3682 | goto exit3; | |
3683 | } | |
f5beed75 | 3684 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3685 | if (error) |
c30dabfe | 3686 | goto exit3; |
f5beed75 | 3687 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3688 | exit3: |
6902d925 DH |
3689 | dput(dentry); |
3690 | exit2: | |
f5beed75 AV |
3691 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
3692 | mnt_drop_write(path.mnt); | |
1da177e4 | 3693 | exit1: |
f5beed75 | 3694 | path_put(&path); |
1da177e4 | 3695 | putname(name); |
c6ee9206 JL |
3696 | if (retry_estale(error, lookup_flags)) { |
3697 | lookup_flags |= LOOKUP_REVAL; | |
3698 | goto retry; | |
3699 | } | |
1da177e4 LT |
3700 | return error; |
3701 | } | |
3702 | ||
3cdad428 | 3703 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3704 | { |
3705 | return do_rmdir(AT_FDCWD, pathname); | |
3706 | } | |
3707 | ||
b21996e3 BF |
3708 | /** |
3709 | * vfs_unlink - unlink a filesystem object | |
3710 | * @dir: parent directory | |
3711 | * @dentry: victim | |
3712 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3713 | * | |
3714 | * The caller must hold dir->i_mutex. | |
3715 | * | |
3716 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3717 | * return a reference to the inode in delegated_inode. The caller | |
3718 | * should then break the delegation on that inode and retry. Because | |
3719 | * breaking a delegation may take a long time, the caller should drop | |
3720 | * dir->i_mutex before doing so. | |
3721 | * | |
3722 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3723 | * be appropriate for callers that expect the underlying filesystem not | |
3724 | * to be NFS exported. | |
3725 | */ | |
3726 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3727 | { |
9accbb97 | 3728 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3729 | int error = may_delete(dir, dentry, 0); |
3730 | ||
3731 | if (error) | |
3732 | return error; | |
3733 | ||
acfa4380 | 3734 | if (!dir->i_op->unlink) |
1da177e4 LT |
3735 | return -EPERM; |
3736 | ||
9accbb97 | 3737 | mutex_lock(&target->i_mutex); |
8ed936b5 | 3738 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3739 | error = -EBUSY; |
3740 | else { | |
3741 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3742 | if (!error) { |
5a14696c BF |
3743 | error = try_break_deleg(target, delegated_inode); |
3744 | if (error) | |
b21996e3 | 3745 | goto out; |
1da177e4 | 3746 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3747 | if (!error) { |
d83c49f3 | 3748 | dont_mount(dentry); |
8ed936b5 EB |
3749 | detach_mounts(dentry); |
3750 | } | |
bec1052e | 3751 | } |
1da177e4 | 3752 | } |
b21996e3 | 3753 | out: |
9accbb97 | 3754 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3755 | |
3756 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3757 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 3758 | fsnotify_link_count(target); |
e234f35c | 3759 | d_delete(dentry); |
1da177e4 | 3760 | } |
0eeca283 | 3761 | |
1da177e4 LT |
3762 | return error; |
3763 | } | |
4d359507 | 3764 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
3765 | |
3766 | /* | |
3767 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3768 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3769 | * writeout happening, and we don't want to prevent access to the directory |
3770 | * while waiting on the I/O. | |
3771 | */ | |
5590ff0d | 3772 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3773 | { |
2ad94ae6 | 3774 | int error; |
91a27b2a | 3775 | struct filename *name; |
1da177e4 | 3776 | struct dentry *dentry; |
f5beed75 AV |
3777 | struct path path; |
3778 | struct qstr last; | |
3779 | int type; | |
1da177e4 | 3780 | struct inode *inode = NULL; |
b21996e3 | 3781 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
3782 | unsigned int lookup_flags = 0; |
3783 | retry: | |
f5beed75 AV |
3784 | name = user_path_parent(dfd, pathname, |
3785 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3786 | if (IS_ERR(name)) |
3787 | return PTR_ERR(name); | |
2ad94ae6 | 3788 | |
1da177e4 | 3789 | error = -EISDIR; |
f5beed75 | 3790 | if (type != LAST_NORM) |
1da177e4 | 3791 | goto exit1; |
0612d9fb | 3792 | |
f5beed75 | 3793 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3794 | if (error) |
3795 | goto exit1; | |
b21996e3 | 3796 | retry_deleg: |
f5beed75 AV |
3797 | mutex_lock_nested(&path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3798 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); | |
1da177e4 LT |
3799 | error = PTR_ERR(dentry); |
3800 | if (!IS_ERR(dentry)) { | |
3801 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 3802 | if (last.name[last.len]) |
50338b88 | 3803 | goto slashes; |
1da177e4 | 3804 | inode = dentry->d_inode; |
b18825a7 | 3805 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
3806 | goto slashes; |
3807 | ihold(inode); | |
f5beed75 | 3808 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 3809 | if (error) |
c30dabfe | 3810 | goto exit2; |
f5beed75 | 3811 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 3812 | exit2: |
1da177e4 LT |
3813 | dput(dentry); |
3814 | } | |
f5beed75 | 3815 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3816 | if (inode) |
3817 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
3818 | inode = NULL; |
3819 | if (delegated_inode) { | |
5a14696c | 3820 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
3821 | if (!error) |
3822 | goto retry_deleg; | |
3823 | } | |
f5beed75 | 3824 | mnt_drop_write(path.mnt); |
1da177e4 | 3825 | exit1: |
f5beed75 | 3826 | path_put(&path); |
1da177e4 | 3827 | putname(name); |
5d18f813 JL |
3828 | if (retry_estale(error, lookup_flags)) { |
3829 | lookup_flags |= LOOKUP_REVAL; | |
3830 | inode = NULL; | |
3831 | goto retry; | |
3832 | } | |
1da177e4 LT |
3833 | return error; |
3834 | ||
3835 | slashes: | |
b18825a7 DH |
3836 | if (d_is_negative(dentry)) |
3837 | error = -ENOENT; | |
44b1d530 | 3838 | else if (d_is_dir(dentry)) |
b18825a7 DH |
3839 | error = -EISDIR; |
3840 | else | |
3841 | error = -ENOTDIR; | |
1da177e4 LT |
3842 | goto exit2; |
3843 | } | |
3844 | ||
2e4d0924 | 3845 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3846 | { |
3847 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3848 | return -EINVAL; | |
3849 | ||
3850 | if (flag & AT_REMOVEDIR) | |
3851 | return do_rmdir(dfd, pathname); | |
3852 | ||
3853 | return do_unlinkat(dfd, pathname); | |
3854 | } | |
3855 | ||
3480b257 | 3856 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3857 | { |
3858 | return do_unlinkat(AT_FDCWD, pathname); | |
3859 | } | |
3860 | ||
db2e747b | 3861 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3862 | { |
a95164d9 | 3863 | int error = may_create(dir, dentry); |
1da177e4 LT |
3864 | |
3865 | if (error) | |
3866 | return error; | |
3867 | ||
acfa4380 | 3868 | if (!dir->i_op->symlink) |
1da177e4 LT |
3869 | return -EPERM; |
3870 | ||
3871 | error = security_inode_symlink(dir, dentry, oldname); | |
3872 | if (error) | |
3873 | return error; | |
3874 | ||
1da177e4 | 3875 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3876 | if (!error) |
f38aa942 | 3877 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3878 | return error; |
3879 | } | |
4d359507 | 3880 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 3881 | |
2e4d0924 HC |
3882 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3883 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3884 | { |
2ad94ae6 | 3885 | int error; |
91a27b2a | 3886 | struct filename *from; |
6902d925 | 3887 | struct dentry *dentry; |
dae6ad8f | 3888 | struct path path; |
f46d3567 | 3889 | unsigned int lookup_flags = 0; |
1da177e4 LT |
3890 | |
3891 | from = getname(oldname); | |
2ad94ae6 | 3892 | if (IS_ERR(from)) |
1da177e4 | 3893 | return PTR_ERR(from); |
f46d3567 JL |
3894 | retry: |
3895 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
3896 | error = PTR_ERR(dentry); |
3897 | if (IS_ERR(dentry)) | |
dae6ad8f | 3898 | goto out_putname; |
6902d925 | 3899 | |
91a27b2a | 3900 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 3901 | if (!error) |
91a27b2a | 3902 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 3903 | done_path_create(&path, dentry); |
f46d3567 JL |
3904 | if (retry_estale(error, lookup_flags)) { |
3905 | lookup_flags |= LOOKUP_REVAL; | |
3906 | goto retry; | |
3907 | } | |
6902d925 | 3908 | out_putname: |
1da177e4 LT |
3909 | putname(from); |
3910 | return error; | |
3911 | } | |
3912 | ||
3480b257 | 3913 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3914 | { |
3915 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3916 | } | |
3917 | ||
146a8595 BF |
3918 | /** |
3919 | * vfs_link - create a new link | |
3920 | * @old_dentry: object to be linked | |
3921 | * @dir: new parent | |
3922 | * @new_dentry: where to create the new link | |
3923 | * @delegated_inode: returns inode needing a delegation break | |
3924 | * | |
3925 | * The caller must hold dir->i_mutex | |
3926 | * | |
3927 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
3928 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
3929 | * inode in delegated_inode. The caller should then break the delegation | |
3930 | * and retry. Because breaking a delegation may take a long time, the | |
3931 | * caller should drop the i_mutex before doing so. | |
3932 | * | |
3933 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3934 | * be appropriate for callers that expect the underlying filesystem not | |
3935 | * to be NFS exported. | |
3936 | */ | |
3937 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
3938 | { |
3939 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 3940 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3941 | int error; |
3942 | ||
3943 | if (!inode) | |
3944 | return -ENOENT; | |
3945 | ||
a95164d9 | 3946 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3947 | if (error) |
3948 | return error; | |
3949 | ||
3950 | if (dir->i_sb != inode->i_sb) | |
3951 | return -EXDEV; | |
3952 | ||
3953 | /* | |
3954 | * A link to an append-only or immutable file cannot be created. | |
3955 | */ | |
3956 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3957 | return -EPERM; | |
acfa4380 | 3958 | if (!dir->i_op->link) |
1da177e4 | 3959 | return -EPERM; |
7e79eedb | 3960 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3961 | return -EPERM; |
3962 | ||
3963 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3964 | if (error) | |
3965 | return error; | |
3966 | ||
7e79eedb | 3967 | mutex_lock(&inode->i_mutex); |
aae8a97d | 3968 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 3969 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 3970 | error = -ENOENT; |
8de52778 AV |
3971 | else if (max_links && inode->i_nlink >= max_links) |
3972 | error = -EMLINK; | |
146a8595 BF |
3973 | else { |
3974 | error = try_break_deleg(inode, delegated_inode); | |
3975 | if (!error) | |
3976 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
3977 | } | |
f4e0c30c AV |
3978 | |
3979 | if (!error && (inode->i_state & I_LINKABLE)) { | |
3980 | spin_lock(&inode->i_lock); | |
3981 | inode->i_state &= ~I_LINKABLE; | |
3982 | spin_unlock(&inode->i_lock); | |
3983 | } | |
7e79eedb | 3984 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3985 | if (!error) |
7e79eedb | 3986 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3987 | return error; |
3988 | } | |
4d359507 | 3989 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
3990 | |
3991 | /* | |
3992 | * Hardlinks are often used in delicate situations. We avoid | |
3993 | * security-related surprises by not following symlinks on the | |
3994 | * newname. --KAB | |
3995 | * | |
3996 | * We don't follow them on the oldname either to be compatible | |
3997 | * with linux 2.0, and to avoid hard-linking to directories | |
3998 | * and other special files. --ADM | |
3999 | */ | |
2e4d0924 HC |
4000 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4001 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
4002 | { |
4003 | struct dentry *new_dentry; | |
dae6ad8f | 4004 | struct path old_path, new_path; |
146a8595 | 4005 | struct inode *delegated_inode = NULL; |
11a7b371 | 4006 | int how = 0; |
1da177e4 | 4007 | int error; |
1da177e4 | 4008 | |
11a7b371 | 4009 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4010 | return -EINVAL; |
11a7b371 | 4011 | /* |
f0cc6ffb LT |
4012 | * To use null names we require CAP_DAC_READ_SEARCH |
4013 | * This ensures that not everyone will be able to create | |
4014 | * handlink using the passed filedescriptor. | |
11a7b371 | 4015 | */ |
f0cc6ffb LT |
4016 | if (flags & AT_EMPTY_PATH) { |
4017 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4018 | return -ENOENT; | |
11a7b371 | 4019 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4020 | } |
11a7b371 AK |
4021 | |
4022 | if (flags & AT_SYMLINK_FOLLOW) | |
4023 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4024 | retry: |
11a7b371 | 4025 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4026 | if (error) |
2ad94ae6 AV |
4027 | return error; |
4028 | ||
442e31ca JL |
4029 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4030 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4031 | error = PTR_ERR(new_dentry); |
6902d925 | 4032 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4033 | goto out; |
4034 | ||
4035 | error = -EXDEV; | |
4036 | if (old_path.mnt != new_path.mnt) | |
4037 | goto out_dput; | |
800179c9 KC |
4038 | error = may_linkat(&old_path); |
4039 | if (unlikely(error)) | |
4040 | goto out_dput; | |
dae6ad8f | 4041 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4042 | if (error) |
a8104a9f | 4043 | goto out_dput; |
146a8595 | 4044 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4045 | out_dput: |
921a1650 | 4046 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4047 | if (delegated_inode) { |
4048 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4049 | if (!error) { |
4050 | path_put(&old_path); | |
146a8595 | 4051 | goto retry; |
d22e6338 | 4052 | } |
146a8595 | 4053 | } |
442e31ca | 4054 | if (retry_estale(error, how)) { |
d22e6338 | 4055 | path_put(&old_path); |
442e31ca JL |
4056 | how |= LOOKUP_REVAL; |
4057 | goto retry; | |
4058 | } | |
1da177e4 | 4059 | out: |
2d8f3038 | 4060 | path_put(&old_path); |
1da177e4 LT |
4061 | |
4062 | return error; | |
4063 | } | |
4064 | ||
3480b257 | 4065 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4066 | { |
c04030e1 | 4067 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4068 | } |
4069 | ||
bc27027a MS |
4070 | /** |
4071 | * vfs_rename - rename a filesystem object | |
4072 | * @old_dir: parent of source | |
4073 | * @old_dentry: source | |
4074 | * @new_dir: parent of destination | |
4075 | * @new_dentry: destination | |
4076 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4077 | * @flags: rename flags |
bc27027a MS |
4078 | * |
4079 | * The caller must hold multiple mutexes--see lock_rename()). | |
4080 | * | |
4081 | * If vfs_rename discovers a delegation in need of breaking at either | |
4082 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4083 | * reference to the inode in delegated_inode. The caller should then | |
4084 | * break the delegation and retry. Because breaking a delegation may | |
4085 | * take a long time, the caller should drop all locks before doing | |
4086 | * so. | |
4087 | * | |
4088 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4089 | * be appropriate for callers that expect the underlying filesystem not | |
4090 | * to be NFS exported. | |
4091 | * | |
1da177e4 LT |
4092 | * The worst of all namespace operations - renaming directory. "Perverted" |
4093 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4094 | * Problems: | |
d03b29a2 | 4095 | * a) we can get into loop creation. |
1da177e4 LT |
4096 | * b) race potential - two innocent renames can create a loop together. |
4097 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4098 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4099 | * story. |
6cedba89 BF |
4100 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4101 | * and source (if it is not a directory). | |
1b1dcc1b | 4102 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4103 | * whether the target exists). Solution: try to be smart with locking |
4104 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4105 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4106 | * move will be locked. Thus we can rank directories by the tree |
4107 | * (ancestors first) and rank all non-directories after them. | |
4108 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4109 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4110 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4111 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4112 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4113 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4114 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4115 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4116 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4117 | * locking]. |
4118 | */ | |
bc27027a MS |
4119 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4120 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4121 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4122 | { |
bc27027a MS |
4123 | int error; |
4124 | bool is_dir = d_is_dir(old_dentry); | |
4125 | const unsigned char *old_name; | |
4126 | struct inode *source = old_dentry->d_inode; | |
9055cba7 | 4127 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4128 | bool new_is_dir = false; |
4129 | unsigned max_links = new_dir->i_sb->s_max_links; | |
bc27027a MS |
4130 | |
4131 | if (source == target) | |
4132 | return 0; | |
4133 | ||
4134 | error = may_delete(old_dir, old_dentry, is_dir); | |
4135 | if (error) | |
4136 | return error; | |
4137 | ||
da1ce067 | 4138 | if (!target) { |
bc27027a | 4139 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4140 | } else { |
4141 | new_is_dir = d_is_dir(new_dentry); | |
4142 | ||
4143 | if (!(flags & RENAME_EXCHANGE)) | |
4144 | error = may_delete(new_dir, new_dentry, is_dir); | |
4145 | else | |
4146 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4147 | } | |
bc27027a MS |
4148 | if (error) |
4149 | return error; | |
4150 | ||
7177a9c4 | 4151 | if (!old_dir->i_op->rename && !old_dir->i_op->rename2) |
bc27027a | 4152 | return -EPERM; |
1da177e4 | 4153 | |
520c8b16 MS |
4154 | if (flags && !old_dir->i_op->rename2) |
4155 | return -EINVAL; | |
4156 | ||
1da177e4 LT |
4157 | /* |
4158 | * If we are going to change the parent - check write permissions, | |
4159 | * we'll need to flip '..'. | |
4160 | */ | |
da1ce067 MS |
4161 | if (new_dir != old_dir) { |
4162 | if (is_dir) { | |
4163 | error = inode_permission(source, MAY_WRITE); | |
4164 | if (error) | |
4165 | return error; | |
4166 | } | |
4167 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4168 | error = inode_permission(target, MAY_WRITE); | |
4169 | if (error) | |
4170 | return error; | |
4171 | } | |
1da177e4 LT |
4172 | } |
4173 | ||
0b3974eb MS |
4174 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4175 | flags); | |
1da177e4 LT |
4176 | if (error) |
4177 | return error; | |
4178 | ||
bc27027a | 4179 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
1d2ef590 | 4180 | dget(new_dentry); |
da1ce067 | 4181 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4182 | lock_two_nondirectories(source, target); |
4183 | else if (target) | |
1b1dcc1b | 4184 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
4185 | |
4186 | error = -EBUSY; | |
7af1364f | 4187 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4188 | goto out; |
4189 | ||
da1ce067 | 4190 | if (max_links && new_dir != old_dir) { |
bc27027a | 4191 | error = -EMLINK; |
da1ce067 | 4192 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4193 | goto out; |
da1ce067 MS |
4194 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4195 | old_dir->i_nlink >= max_links) | |
4196 | goto out; | |
4197 | } | |
4198 | if (is_dir && !(flags & RENAME_EXCHANGE) && target) | |
4199 | shrink_dcache_parent(new_dentry); | |
4200 | if (!is_dir) { | |
bc27027a | 4201 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4202 | if (error) |
4203 | goto out; | |
da1ce067 MS |
4204 | } |
4205 | if (target && !new_is_dir) { | |
4206 | error = try_break_deleg(target, delegated_inode); | |
4207 | if (error) | |
4208 | goto out; | |
8e6d782c | 4209 | } |
7177a9c4 | 4210 | if (!old_dir->i_op->rename2) { |
520c8b16 MS |
4211 | error = old_dir->i_op->rename(old_dir, old_dentry, |
4212 | new_dir, new_dentry); | |
4213 | } else { | |
7177a9c4 | 4214 | WARN_ON(old_dir->i_op->rename != NULL); |
520c8b16 MS |
4215 | error = old_dir->i_op->rename2(old_dir, old_dentry, |
4216 | new_dir, new_dentry, flags); | |
4217 | } | |
51892bbb SW |
4218 | if (error) |
4219 | goto out; | |
4220 | ||
da1ce067 | 4221 | if (!(flags & RENAME_EXCHANGE) && target) { |
bc27027a MS |
4222 | if (is_dir) |
4223 | target->i_flags |= S_DEAD; | |
51892bbb | 4224 | dont_mount(new_dentry); |
8ed936b5 | 4225 | detach_mounts(new_dentry); |
bc27027a | 4226 | } |
da1ce067 MS |
4227 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4228 | if (!(flags & RENAME_EXCHANGE)) | |
4229 | d_move(old_dentry, new_dentry); | |
4230 | else | |
4231 | d_exchange(old_dentry, new_dentry); | |
4232 | } | |
51892bbb | 4233 | out: |
da1ce067 | 4234 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4235 | unlock_two_nondirectories(source, target); |
4236 | else if (target) | |
4237 | mutex_unlock(&target->i_mutex); | |
1da177e4 | 4238 | dput(new_dentry); |
da1ce067 | 4239 | if (!error) { |
123df294 | 4240 | fsnotify_move(old_dir, new_dir, old_name, is_dir, |
da1ce067 MS |
4241 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4242 | if (flags & RENAME_EXCHANGE) { | |
4243 | fsnotify_move(new_dir, old_dir, old_dentry->d_name.name, | |
4244 | new_is_dir, NULL, new_dentry); | |
4245 | } | |
4246 | } | |
0eeca283 RL |
4247 | fsnotify_oldname_free(old_name); |
4248 | ||
1da177e4 LT |
4249 | return error; |
4250 | } | |
4d359507 | 4251 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4252 | |
520c8b16 MS |
4253 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4254 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
1da177e4 | 4255 | { |
2ad94ae6 AV |
4256 | struct dentry *old_dentry, *new_dentry; |
4257 | struct dentry *trap; | |
f5beed75 AV |
4258 | struct path old_path, new_path; |
4259 | struct qstr old_last, new_last; | |
4260 | int old_type, new_type; | |
8e6d782c | 4261 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4262 | struct filename *from; |
4263 | struct filename *to; | |
f5beed75 | 4264 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4265 | bool should_retry = false; |
2ad94ae6 | 4266 | int error; |
520c8b16 | 4267 | |
0d7a8555 | 4268 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4269 | return -EINVAL; |
4270 | ||
0d7a8555 MS |
4271 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4272 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4273 | return -EINVAL; |
4274 | ||
0d7a8555 MS |
4275 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4276 | return -EPERM; | |
4277 | ||
f5beed75 AV |
4278 | if (flags & RENAME_EXCHANGE) |
4279 | target_flags = 0; | |
4280 | ||
c6a94284 | 4281 | retry: |
f5beed75 AV |
4282 | from = user_path_parent(olddfd, oldname, |
4283 | &old_path, &old_last, &old_type, lookup_flags); | |
91a27b2a JL |
4284 | if (IS_ERR(from)) { |
4285 | error = PTR_ERR(from); | |
1da177e4 | 4286 | goto exit; |
91a27b2a | 4287 | } |
1da177e4 | 4288 | |
f5beed75 AV |
4289 | to = user_path_parent(newdfd, newname, |
4290 | &new_path, &new_last, &new_type, lookup_flags); | |
91a27b2a JL |
4291 | if (IS_ERR(to)) { |
4292 | error = PTR_ERR(to); | |
1da177e4 | 4293 | goto exit1; |
91a27b2a | 4294 | } |
1da177e4 LT |
4295 | |
4296 | error = -EXDEV; | |
f5beed75 | 4297 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4298 | goto exit2; |
4299 | ||
1da177e4 | 4300 | error = -EBUSY; |
f5beed75 | 4301 | if (old_type != LAST_NORM) |
1da177e4 LT |
4302 | goto exit2; |
4303 | ||
0a7c3937 MS |
4304 | if (flags & RENAME_NOREPLACE) |
4305 | error = -EEXIST; | |
f5beed75 | 4306 | if (new_type != LAST_NORM) |
1da177e4 LT |
4307 | goto exit2; |
4308 | ||
f5beed75 | 4309 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4310 | if (error) |
4311 | goto exit2; | |
4312 | ||
8e6d782c | 4313 | retry_deleg: |
f5beed75 | 4314 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4315 | |
f5beed75 | 4316 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4317 | error = PTR_ERR(old_dentry); |
4318 | if (IS_ERR(old_dentry)) | |
4319 | goto exit3; | |
4320 | /* source must exist */ | |
4321 | error = -ENOENT; | |
b18825a7 | 4322 | if (d_is_negative(old_dentry)) |
1da177e4 | 4323 | goto exit4; |
f5beed75 | 4324 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4325 | error = PTR_ERR(new_dentry); |
4326 | if (IS_ERR(new_dentry)) | |
4327 | goto exit4; | |
4328 | error = -EEXIST; | |
4329 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4330 | goto exit5; | |
da1ce067 MS |
4331 | if (flags & RENAME_EXCHANGE) { |
4332 | error = -ENOENT; | |
4333 | if (d_is_negative(new_dentry)) | |
4334 | goto exit5; | |
4335 | ||
4336 | if (!d_is_dir(new_dentry)) { | |
4337 | error = -ENOTDIR; | |
f5beed75 | 4338 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4339 | goto exit5; |
4340 | } | |
4341 | } | |
1da177e4 | 4342 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4343 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4344 | error = -ENOTDIR; |
f5beed75 | 4345 | if (old_last.name[old_last.len]) |
0a7c3937 | 4346 | goto exit5; |
f5beed75 | 4347 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4348 | goto exit5; |
1da177e4 LT |
4349 | } |
4350 | /* source should not be ancestor of target */ | |
4351 | error = -EINVAL; | |
4352 | if (old_dentry == trap) | |
0a7c3937 | 4353 | goto exit5; |
1da177e4 | 4354 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4355 | if (!(flags & RENAME_EXCHANGE)) |
4356 | error = -ENOTEMPTY; | |
1da177e4 LT |
4357 | if (new_dentry == trap) |
4358 | goto exit5; | |
4359 | ||
f5beed75 AV |
4360 | error = security_path_rename(&old_path, old_dentry, |
4361 | &new_path, new_dentry, flags); | |
be6d3e56 | 4362 | if (error) |
c30dabfe | 4363 | goto exit5; |
f5beed75 AV |
4364 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4365 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4366 | &delegated_inode, flags); |
1da177e4 LT |
4367 | exit5: |
4368 | dput(new_dentry); | |
4369 | exit4: | |
4370 | dput(old_dentry); | |
4371 | exit3: | |
f5beed75 | 4372 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4373 | if (delegated_inode) { |
4374 | error = break_deleg_wait(&delegated_inode); | |
4375 | if (!error) | |
4376 | goto retry_deleg; | |
4377 | } | |
f5beed75 | 4378 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4379 | exit2: |
c6a94284 JL |
4380 | if (retry_estale(error, lookup_flags)) |
4381 | should_retry = true; | |
f5beed75 | 4382 | path_put(&new_path); |
2ad94ae6 | 4383 | putname(to); |
1da177e4 | 4384 | exit1: |
f5beed75 | 4385 | path_put(&old_path); |
1da177e4 | 4386 | putname(from); |
c6a94284 JL |
4387 | if (should_retry) { |
4388 | should_retry = false; | |
4389 | lookup_flags |= LOOKUP_REVAL; | |
4390 | goto retry; | |
4391 | } | |
2ad94ae6 | 4392 | exit: |
1da177e4 LT |
4393 | return error; |
4394 | } | |
4395 | ||
520c8b16 MS |
4396 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4397 | int, newdfd, const char __user *, newname) | |
4398 | { | |
4399 | return sys_renameat2(olddfd, oldname, newdfd, newname, 0); | |
4400 | } | |
4401 | ||
a26eab24 | 4402 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4403 | { |
520c8b16 | 4404 | return sys_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4405 | } |
4406 | ||
787fb6bc MS |
4407 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4408 | { | |
4409 | int error = may_create(dir, dentry); | |
4410 | if (error) | |
4411 | return error; | |
4412 | ||
4413 | if (!dir->i_op->mknod) | |
4414 | return -EPERM; | |
4415 | ||
4416 | return dir->i_op->mknod(dir, dentry, | |
4417 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4418 | } | |
4419 | EXPORT_SYMBOL(vfs_whiteout); | |
4420 | ||
5d826c84 | 4421 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4422 | { |
5d826c84 | 4423 | int len = PTR_ERR(link); |
1da177e4 LT |
4424 | if (IS_ERR(link)) |
4425 | goto out; | |
4426 | ||
4427 | len = strlen(link); | |
4428 | if (len > (unsigned) buflen) | |
4429 | len = buflen; | |
4430 | if (copy_to_user(buffer, link, len)) | |
4431 | len = -EFAULT; | |
4432 | out: | |
4433 | return len; | |
4434 | } | |
5d826c84 | 4435 | EXPORT_SYMBOL(readlink_copy); |
1da177e4 LT |
4436 | |
4437 | /* | |
4438 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
4439 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
4440 | * using) it for any given inode is up to filesystem. | |
4441 | */ | |
4442 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4443 | { | |
cc314eef | 4444 | void *cookie; |
d4dee48b | 4445 | const char *link = dentry->d_inode->i_link; |
694a1764 | 4446 | int res; |
cc314eef | 4447 | |
d4dee48b AV |
4448 | if (!link) { |
4449 | link = dentry->d_inode->i_op->follow_link(dentry, &cookie, NULL); | |
4450 | if (IS_ERR(link)) | |
4451 | return PTR_ERR(link); | |
4452 | } | |
680baacb AV |
4453 | res = readlink_copy(buffer, buflen, link); |
4454 | if (cookie && dentry->d_inode->i_op->put_link) | |
4455 | dentry->d_inode->i_op->put_link(dentry, cookie); | |
694a1764 | 4456 | return res; |
1da177e4 | 4457 | } |
4d359507 | 4458 | EXPORT_SYMBOL(generic_readlink); |
1da177e4 | 4459 | |
1da177e4 LT |
4460 | /* get the link contents into pagecache */ |
4461 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
4462 | { | |
ebd09abb DG |
4463 | char *kaddr; |
4464 | struct page *page; | |
1da177e4 | 4465 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 4466 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 4467 | if (IS_ERR(page)) |
6fe6900e | 4468 | return (char*)page; |
1da177e4 | 4469 | *ppage = page; |
ebd09abb DG |
4470 | kaddr = kmap(page); |
4471 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
4472 | return kaddr; | |
1da177e4 LT |
4473 | } |
4474 | ||
4475 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4476 | { | |
4477 | struct page *page = NULL; | |
5d826c84 | 4478 | int res = readlink_copy(buffer, buflen, page_getlink(dentry, &page)); |
1da177e4 LT |
4479 | if (page) { |
4480 | kunmap(page); | |
4481 | page_cache_release(page); | |
4482 | } | |
4483 | return res; | |
4484 | } | |
4d359507 | 4485 | EXPORT_SYMBOL(page_readlink); |
1da177e4 | 4486 | |
680baacb | 4487 | const char *page_follow_link_light(struct dentry *dentry, void **cookie, struct nameidata *nd) |
1da177e4 | 4488 | { |
cc314eef | 4489 | struct page *page = NULL; |
680baacb AV |
4490 | char *res = page_getlink(dentry, &page); |
4491 | if (!IS_ERR(res)) | |
4492 | *cookie = page; | |
4493 | return res; | |
1da177e4 | 4494 | } |
4d359507 | 4495 | EXPORT_SYMBOL(page_follow_link_light); |
1da177e4 | 4496 | |
680baacb | 4497 | void page_put_link(struct dentry *dentry, void *cookie) |
1da177e4 | 4498 | { |
cc314eef | 4499 | struct page *page = cookie; |
680baacb AV |
4500 | kunmap(page); |
4501 | page_cache_release(page); | |
1da177e4 | 4502 | } |
4d359507 | 4503 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4504 | |
54566b2c NP |
4505 | /* |
4506 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4507 | */ | |
4508 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4509 | { |
4510 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4511 | struct page *page; |
afddba49 | 4512 | void *fsdata; |
beb497ab | 4513 | int err; |
1da177e4 | 4514 | char *kaddr; |
54566b2c NP |
4515 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
4516 | if (nofs) | |
4517 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4518 | |
7e53cac4 | 4519 | retry: |
afddba49 | 4520 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4521 | flags, &page, &fsdata); |
1da177e4 | 4522 | if (err) |
afddba49 NP |
4523 | goto fail; |
4524 | ||
e8e3c3d6 | 4525 | kaddr = kmap_atomic(page); |
1da177e4 | 4526 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 4527 | kunmap_atomic(kaddr); |
afddba49 NP |
4528 | |
4529 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4530 | page, fsdata); | |
1da177e4 LT |
4531 | if (err < 0) |
4532 | goto fail; | |
afddba49 NP |
4533 | if (err < len-1) |
4534 | goto retry; | |
4535 | ||
1da177e4 LT |
4536 | mark_inode_dirty(inode); |
4537 | return 0; | |
1da177e4 LT |
4538 | fail: |
4539 | return err; | |
4540 | } | |
4d359507 | 4541 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4542 | |
0adb25d2 KK |
4543 | int page_symlink(struct inode *inode, const char *symname, int len) |
4544 | { | |
4545 | return __page_symlink(inode, symname, len, | |
54566b2c | 4546 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 | 4547 | } |
4d359507 | 4548 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4549 | |
92e1d5be | 4550 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
4551 | .readlink = generic_readlink, |
4552 | .follow_link = page_follow_link_light, | |
4553 | .put_link = page_put_link, | |
4554 | }; | |
1da177e4 | 4555 | EXPORT_SYMBOL(page_symlink_inode_operations); |