]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
630d9c47 | 18 | #include <linux/export.h> |
44696908 | 19 | #include <linux/kernel.h> |
1da177e4 LT |
20 | #include <linux/slab.h> |
21 | #include <linux/fs.h> | |
22 | #include <linux/namei.h> | |
1da177e4 | 23 | #include <linux/pagemap.h> |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
6146f0d5 | 27 | #include <linux/ima.h> |
1da177e4 LT |
28 | #include <linux/syscalls.h> |
29 | #include <linux/mount.h> | |
30 | #include <linux/audit.h> | |
16f7e0fe | 31 | #include <linux/capability.h> |
834f2a4a | 32 | #include <linux/file.h> |
5590ff0d | 33 | #include <linux/fcntl.h> |
08ce5f16 | 34 | #include <linux/device_cgroup.h> |
5ad4e53b | 35 | #include <linux/fs_struct.h> |
e77819e5 | 36 | #include <linux/posix_acl.h> |
99d263d4 | 37 | #include <linux/hash.h> |
1da177e4 LT |
38 | #include <asm/uaccess.h> |
39 | ||
e81e3f4d | 40 | #include "internal.h" |
c7105365 | 41 | #include "mount.h" |
e81e3f4d | 42 | |
1da177e4 LT |
43 | /* [Feb-1997 T. Schoebel-Theuer] |
44 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
45 | * were necessary because of omirr. The reason is that omirr needs | |
46 | * to know the _real_ pathname, not the user-supplied one, in case | |
47 | * of symlinks (and also when transname replacements occur). | |
48 | * | |
49 | * The new code replaces the old recursive symlink resolution with | |
50 | * an iterative one (in case of non-nested symlink chains). It does | |
51 | * this with calls to <fs>_follow_link(). | |
52 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
53 | * replaced with a single function lookup_dentry() that can handle all | |
54 | * the special cases of the former code. | |
55 | * | |
56 | * With the new dcache, the pathname is stored at each inode, at least as | |
57 | * long as the refcount of the inode is positive. As a side effect, the | |
58 | * size of the dcache depends on the inode cache and thus is dynamic. | |
59 | * | |
60 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
61 | * resolution to correspond with current state of the code. | |
62 | * | |
63 | * Note that the symlink resolution is not *completely* iterative. | |
64 | * There is still a significant amount of tail- and mid- recursion in | |
65 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
66 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
67 | * may return different results than <fs>_follow_link(). Many virtual | |
68 | * filesystems (including /proc) exhibit this behavior. | |
69 | */ | |
70 | ||
71 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
72 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
73 | * and the name already exists in form of a symlink, try to create the new | |
74 | * name indicated by the symlink. The old code always complained that the | |
75 | * name already exists, due to not following the symlink even if its target | |
76 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 77 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
78 | * |
79 | * I don't know which semantics is the right one, since I have no access | |
80 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
81 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
82 | * "old" one. Personally, I think the new semantics is much more logical. | |
83 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
84 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
85 | * and in the old Linux semantics. | |
86 | */ | |
87 | ||
88 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
89 | * semantics. See the comments in "open_namei" and "do_link" below. | |
90 | * | |
91 | * [10-Sep-98 Alan Modra] Another symlink change. | |
92 | */ | |
93 | ||
94 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
95 | * inside the path - always follow. | |
96 | * in the last component in creation/removal/renaming - never follow. | |
97 | * if LOOKUP_FOLLOW passed - follow. | |
98 | * if the pathname has trailing slashes - follow. | |
99 | * otherwise - don't follow. | |
100 | * (applied in that order). | |
101 | * | |
102 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
103 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
104 | * During the 2.4 we need to fix the userland stuff depending on it - | |
105 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
106 | * XEmacs seems to be relying on it... | |
107 | */ | |
108 | /* | |
109 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 110 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
111 | * any extra contention... |
112 | */ | |
113 | ||
114 | /* In order to reduce some races, while at the same time doing additional | |
115 | * checking and hopefully speeding things up, we copy filenames to the | |
116 | * kernel data space before using them.. | |
117 | * | |
118 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
119 | * PATH_MAX includes the nul terminator --RR. | |
120 | */ | |
91a27b2a | 121 | |
fd2f7cb5 | 122 | #define EMBEDDED_NAME_MAX (PATH_MAX - offsetof(struct filename, iname)) |
7950e385 | 123 | |
51f39a1f | 124 | struct filename * |
91a27b2a JL |
125 | getname_flags(const char __user *filename, int flags, int *empty) |
126 | { | |
94b5d262 | 127 | struct filename *result; |
7950e385 | 128 | char *kname; |
94b5d262 | 129 | int len; |
4043cde8 | 130 | |
7ac86265 JL |
131 | result = audit_reusename(filename); |
132 | if (result) | |
133 | return result; | |
134 | ||
7950e385 | 135 | result = __getname(); |
3f9f0aa6 | 136 | if (unlikely(!result)) |
4043cde8 EP |
137 | return ERR_PTR(-ENOMEM); |
138 | ||
7950e385 JL |
139 | /* |
140 | * First, try to embed the struct filename inside the names_cache | |
141 | * allocation | |
142 | */ | |
fd2f7cb5 | 143 | kname = (char *)result->iname; |
91a27b2a | 144 | result->name = kname; |
7950e385 | 145 | |
94b5d262 | 146 | len = strncpy_from_user(kname, filename, EMBEDDED_NAME_MAX); |
91a27b2a | 147 | if (unlikely(len < 0)) { |
94b5d262 AV |
148 | __putname(result); |
149 | return ERR_PTR(len); | |
91a27b2a | 150 | } |
3f9f0aa6 | 151 | |
7950e385 JL |
152 | /* |
153 | * Uh-oh. We have a name that's approaching PATH_MAX. Allocate a | |
154 | * separate struct filename so we can dedicate the entire | |
155 | * names_cache allocation for the pathname, and re-do the copy from | |
156 | * userland. | |
157 | */ | |
94b5d262 | 158 | if (unlikely(len == EMBEDDED_NAME_MAX)) { |
fd2f7cb5 | 159 | const size_t size = offsetof(struct filename, iname[1]); |
7950e385 JL |
160 | kname = (char *)result; |
161 | ||
fd2f7cb5 AV |
162 | /* |
163 | * size is chosen that way we to guarantee that | |
164 | * result->iname[0] is within the same object and that | |
165 | * kname can't be equal to result->iname, no matter what. | |
166 | */ | |
167 | result = kzalloc(size, GFP_KERNEL); | |
94b5d262 AV |
168 | if (unlikely(!result)) { |
169 | __putname(kname); | |
170 | return ERR_PTR(-ENOMEM); | |
7950e385 JL |
171 | } |
172 | result->name = kname; | |
94b5d262 AV |
173 | len = strncpy_from_user(kname, filename, PATH_MAX); |
174 | if (unlikely(len < 0)) { | |
175 | __putname(kname); | |
176 | kfree(result); | |
177 | return ERR_PTR(len); | |
178 | } | |
179 | if (unlikely(len == PATH_MAX)) { | |
180 | __putname(kname); | |
181 | kfree(result); | |
182 | return ERR_PTR(-ENAMETOOLONG); | |
183 | } | |
7950e385 JL |
184 | } |
185 | ||
94b5d262 | 186 | result->refcnt = 1; |
3f9f0aa6 LT |
187 | /* The empty path is special. */ |
188 | if (unlikely(!len)) { | |
189 | if (empty) | |
4043cde8 | 190 | *empty = 1; |
94b5d262 AV |
191 | if (!(flags & LOOKUP_EMPTY)) { |
192 | putname(result); | |
193 | return ERR_PTR(-ENOENT); | |
194 | } | |
1da177e4 | 195 | } |
3f9f0aa6 | 196 | |
7950e385 | 197 | result->uptr = filename; |
c4ad8f98 | 198 | result->aname = NULL; |
7950e385 JL |
199 | audit_getname(result); |
200 | return result; | |
1da177e4 LT |
201 | } |
202 | ||
91a27b2a JL |
203 | struct filename * |
204 | getname(const char __user * filename) | |
f52e0c11 | 205 | { |
f7493e5d | 206 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
207 | } |
208 | ||
c4ad8f98 LT |
209 | struct filename * |
210 | getname_kernel(const char * filename) | |
211 | { | |
212 | struct filename *result; | |
08518549 | 213 | int len = strlen(filename) + 1; |
c4ad8f98 LT |
214 | |
215 | result = __getname(); | |
216 | if (unlikely(!result)) | |
217 | return ERR_PTR(-ENOMEM); | |
218 | ||
08518549 | 219 | if (len <= EMBEDDED_NAME_MAX) { |
fd2f7cb5 | 220 | result->name = (char *)result->iname; |
08518549 PM |
221 | } else if (len <= PATH_MAX) { |
222 | struct filename *tmp; | |
223 | ||
224 | tmp = kmalloc(sizeof(*tmp), GFP_KERNEL); | |
225 | if (unlikely(!tmp)) { | |
226 | __putname(result); | |
227 | return ERR_PTR(-ENOMEM); | |
228 | } | |
229 | tmp->name = (char *)result; | |
08518549 PM |
230 | result = tmp; |
231 | } else { | |
232 | __putname(result); | |
233 | return ERR_PTR(-ENAMETOOLONG); | |
234 | } | |
235 | memcpy((char *)result->name, filename, len); | |
c4ad8f98 LT |
236 | result->uptr = NULL; |
237 | result->aname = NULL; | |
55422d0b | 238 | result->refcnt = 1; |
fd3522fd | 239 | audit_getname(result); |
c4ad8f98 | 240 | |
c4ad8f98 LT |
241 | return result; |
242 | } | |
243 | ||
91a27b2a | 244 | void putname(struct filename *name) |
1da177e4 | 245 | { |
55422d0b PM |
246 | BUG_ON(name->refcnt <= 0); |
247 | ||
248 | if (--name->refcnt > 0) | |
249 | return; | |
250 | ||
fd2f7cb5 | 251 | if (name->name != name->iname) { |
55422d0b PM |
252 | __putname(name->name); |
253 | kfree(name); | |
254 | } else | |
255 | __putname(name); | |
1da177e4 | 256 | } |
1da177e4 | 257 | |
e77819e5 LT |
258 | static int check_acl(struct inode *inode, int mask) |
259 | { | |
84635d68 | 260 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
261 | struct posix_acl *acl; |
262 | ||
e77819e5 | 263 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
264 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
265 | if (!acl) | |
e77819e5 | 266 | return -EAGAIN; |
3567866b AV |
267 | /* no ->get_acl() calls in RCU mode... */ |
268 | if (acl == ACL_NOT_CACHED) | |
269 | return -ECHILD; | |
206b1d09 | 270 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
271 | } |
272 | ||
2982baa2 CH |
273 | acl = get_acl(inode, ACL_TYPE_ACCESS); |
274 | if (IS_ERR(acl)) | |
275 | return PTR_ERR(acl); | |
e77819e5 LT |
276 | if (acl) { |
277 | int error = posix_acl_permission(inode, acl, mask); | |
278 | posix_acl_release(acl); | |
279 | return error; | |
280 | } | |
84635d68 | 281 | #endif |
e77819e5 LT |
282 | |
283 | return -EAGAIN; | |
284 | } | |
285 | ||
5909ccaa | 286 | /* |
948409c7 | 287 | * This does the basic permission checking |
1da177e4 | 288 | */ |
7e40145e | 289 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 290 | { |
26cf46be | 291 | unsigned int mode = inode->i_mode; |
1da177e4 | 292 | |
8e96e3b7 | 293 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
294 | mode >>= 6; |
295 | else { | |
e77819e5 | 296 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 297 | int error = check_acl(inode, mask); |
b74c79e9 NP |
298 | if (error != -EAGAIN) |
299 | return error; | |
1da177e4 LT |
300 | } |
301 | ||
302 | if (in_group_p(inode->i_gid)) | |
303 | mode >>= 3; | |
304 | } | |
305 | ||
306 | /* | |
307 | * If the DACs are ok we don't need any capability check. | |
308 | */ | |
9c2c7039 | 309 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 310 | return 0; |
5909ccaa LT |
311 | return -EACCES; |
312 | } | |
313 | ||
314 | /** | |
b74c79e9 | 315 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 316 | * @inode: inode to check access rights for |
8fd90c8d | 317 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
318 | * |
319 | * Used to check for read/write/execute permissions on a file. | |
320 | * We use "fsuid" for this, letting us set arbitrary permissions | |
321 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
322 | * are used for other things. |
323 | * | |
324 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
325 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
326 | * It would then be called again in ref-walk mode. | |
5909ccaa | 327 | */ |
2830ba7f | 328 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
329 | { |
330 | int ret; | |
331 | ||
332 | /* | |
948409c7 | 333 | * Do the basic permission checks. |
5909ccaa | 334 | */ |
7e40145e | 335 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
336 | if (ret != -EACCES) |
337 | return ret; | |
1da177e4 | 338 | |
d594e7ec AV |
339 | if (S_ISDIR(inode->i_mode)) { |
340 | /* DACs are overridable for directories */ | |
23adbe12 | 341 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) |
d594e7ec AV |
342 | return 0; |
343 | if (!(mask & MAY_WRITE)) | |
23adbe12 AL |
344 | if (capable_wrt_inode_uidgid(inode, |
345 | CAP_DAC_READ_SEARCH)) | |
d594e7ec AV |
346 | return 0; |
347 | return -EACCES; | |
348 | } | |
1da177e4 LT |
349 | /* |
350 | * Read/write DACs are always overridable. | |
d594e7ec AV |
351 | * Executable DACs are overridable when there is |
352 | * at least one exec bit set. | |
1da177e4 | 353 | */ |
d594e7ec | 354 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
23adbe12 | 355 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) |
1da177e4 LT |
356 | return 0; |
357 | ||
358 | /* | |
359 | * Searching includes executable on directories, else just read. | |
360 | */ | |
7ea66001 | 361 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 362 | if (mask == MAY_READ) |
23adbe12 | 363 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
364 | return 0; |
365 | ||
366 | return -EACCES; | |
367 | } | |
4d359507 | 368 | EXPORT_SYMBOL(generic_permission); |
1da177e4 | 369 | |
3ddcd056 LT |
370 | /* |
371 | * We _really_ want to just do "generic_permission()" without | |
372 | * even looking at the inode->i_op values. So we keep a cache | |
373 | * flag in inode->i_opflags, that says "this has not special | |
374 | * permission function, use the fast case". | |
375 | */ | |
376 | static inline int do_inode_permission(struct inode *inode, int mask) | |
377 | { | |
378 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
379 | if (likely(inode->i_op->permission)) | |
380 | return inode->i_op->permission(inode, mask); | |
381 | ||
382 | /* This gets set once for the inode lifetime */ | |
383 | spin_lock(&inode->i_lock); | |
384 | inode->i_opflags |= IOP_FASTPERM; | |
385 | spin_unlock(&inode->i_lock); | |
386 | } | |
387 | return generic_permission(inode, mask); | |
388 | } | |
389 | ||
cb23beb5 | 390 | /** |
0bdaea90 DH |
391 | * __inode_permission - Check for access rights to a given inode |
392 | * @inode: Inode to check permission on | |
393 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
cb23beb5 | 394 | * |
0bdaea90 | 395 | * Check for read/write/execute permissions on an inode. |
948409c7 AG |
396 | * |
397 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
0bdaea90 DH |
398 | * |
399 | * This does not check for a read-only file system. You probably want | |
400 | * inode_permission(). | |
cb23beb5 | 401 | */ |
0bdaea90 | 402 | int __inode_permission(struct inode *inode, int mask) |
1da177e4 | 403 | { |
e6305c43 | 404 | int retval; |
1da177e4 | 405 | |
3ddcd056 | 406 | if (unlikely(mask & MAY_WRITE)) { |
1da177e4 LT |
407 | /* |
408 | * Nobody gets write access to an immutable file. | |
409 | */ | |
410 | if (IS_IMMUTABLE(inode)) | |
411 | return -EACCES; | |
412 | } | |
413 | ||
3ddcd056 | 414 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
415 | if (retval) |
416 | return retval; | |
417 | ||
08ce5f16 SH |
418 | retval = devcgroup_inode_permission(inode, mask); |
419 | if (retval) | |
420 | return retval; | |
421 | ||
d09ca739 | 422 | return security_inode_permission(inode, mask); |
1da177e4 | 423 | } |
bd5d0856 | 424 | EXPORT_SYMBOL(__inode_permission); |
1da177e4 | 425 | |
0bdaea90 DH |
426 | /** |
427 | * sb_permission - Check superblock-level permissions | |
428 | * @sb: Superblock of inode to check permission on | |
55852635 | 429 | * @inode: Inode to check permission on |
0bdaea90 DH |
430 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
431 | * | |
432 | * Separate out file-system wide checks from inode-specific permission checks. | |
433 | */ | |
434 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
435 | { | |
436 | if (unlikely(mask & MAY_WRITE)) { | |
437 | umode_t mode = inode->i_mode; | |
438 | ||
439 | /* Nobody gets write access to a read-only fs. */ | |
440 | if ((sb->s_flags & MS_RDONLY) && | |
441 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
442 | return -EROFS; | |
443 | } | |
444 | return 0; | |
445 | } | |
446 | ||
447 | /** | |
448 | * inode_permission - Check for access rights to a given inode | |
449 | * @inode: Inode to check permission on | |
450 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
451 | * | |
452 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
453 | * this, letting us set arbitrary permissions for filesystem access without | |
454 | * changing the "normal" UIDs which are used for other things. | |
455 | * | |
456 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
457 | */ | |
458 | int inode_permission(struct inode *inode, int mask) | |
459 | { | |
460 | int retval; | |
461 | ||
462 | retval = sb_permission(inode->i_sb, inode, mask); | |
463 | if (retval) | |
464 | return retval; | |
465 | return __inode_permission(inode, mask); | |
466 | } | |
4d359507 | 467 | EXPORT_SYMBOL(inode_permission); |
0bdaea90 | 468 | |
5dd784d0 JB |
469 | /** |
470 | * path_get - get a reference to a path | |
471 | * @path: path to get the reference to | |
472 | * | |
473 | * Given a path increment the reference count to the dentry and the vfsmount. | |
474 | */ | |
dcf787f3 | 475 | void path_get(const struct path *path) |
5dd784d0 JB |
476 | { |
477 | mntget(path->mnt); | |
478 | dget(path->dentry); | |
479 | } | |
480 | EXPORT_SYMBOL(path_get); | |
481 | ||
1d957f9b JB |
482 | /** |
483 | * path_put - put a reference to a path | |
484 | * @path: path to put the reference to | |
485 | * | |
486 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
487 | */ | |
dcf787f3 | 488 | void path_put(const struct path *path) |
1da177e4 | 489 | { |
1d957f9b JB |
490 | dput(path->dentry); |
491 | mntput(path->mnt); | |
1da177e4 | 492 | } |
1d957f9b | 493 | EXPORT_SYMBOL(path_put); |
1da177e4 | 494 | |
894bc8c4 | 495 | #define EMBEDDED_LEVELS 2 |
1f55a6ec AV |
496 | struct nameidata { |
497 | struct path path; | |
1cf2665b | 498 | struct qstr last; |
1f55a6ec AV |
499 | struct path root; |
500 | struct inode *inode; /* path.dentry.d_inode */ | |
501 | unsigned int flags; | |
9883d185 | 502 | unsigned seq, m_seq; |
1f55a6ec AV |
503 | int last_type; |
504 | unsigned depth; | |
756daf26 | 505 | int total_link_count; |
697fc6ca AV |
506 | struct saved { |
507 | struct path link; | |
fceef393 | 508 | struct delayed_call done; |
697fc6ca | 509 | const char *name; |
0450b2d1 | 510 | unsigned seq; |
894bc8c4 | 511 | } *stack, internal[EMBEDDED_LEVELS]; |
9883d185 AV |
512 | struct filename *name; |
513 | struct nameidata *saved; | |
fceef393 | 514 | struct inode *link_inode; |
9883d185 AV |
515 | unsigned root_seq; |
516 | int dfd; | |
1f55a6ec AV |
517 | }; |
518 | ||
9883d185 | 519 | static void set_nameidata(struct nameidata *p, int dfd, struct filename *name) |
894bc8c4 | 520 | { |
756daf26 N |
521 | struct nameidata *old = current->nameidata; |
522 | p->stack = p->internal; | |
c8a53ee5 AV |
523 | p->dfd = dfd; |
524 | p->name = name; | |
756daf26 | 525 | p->total_link_count = old ? old->total_link_count : 0; |
9883d185 | 526 | p->saved = old; |
756daf26 | 527 | current->nameidata = p; |
894bc8c4 AV |
528 | } |
529 | ||
9883d185 | 530 | static void restore_nameidata(void) |
894bc8c4 | 531 | { |
9883d185 | 532 | struct nameidata *now = current->nameidata, *old = now->saved; |
756daf26 N |
533 | |
534 | current->nameidata = old; | |
535 | if (old) | |
536 | old->total_link_count = now->total_link_count; | |
e1a63bbc | 537 | if (now->stack != now->internal) |
756daf26 | 538 | kfree(now->stack); |
894bc8c4 AV |
539 | } |
540 | ||
541 | static int __nd_alloc_stack(struct nameidata *nd) | |
542 | { | |
bc40aee0 AV |
543 | struct saved *p; |
544 | ||
545 | if (nd->flags & LOOKUP_RCU) { | |
546 | p= kmalloc(MAXSYMLINKS * sizeof(struct saved), | |
547 | GFP_ATOMIC); | |
548 | if (unlikely(!p)) | |
549 | return -ECHILD; | |
550 | } else { | |
551 | p= kmalloc(MAXSYMLINKS * sizeof(struct saved), | |
894bc8c4 | 552 | GFP_KERNEL); |
bc40aee0 AV |
553 | if (unlikely(!p)) |
554 | return -ENOMEM; | |
555 | } | |
894bc8c4 AV |
556 | memcpy(p, nd->internal, sizeof(nd->internal)); |
557 | nd->stack = p; | |
558 | return 0; | |
559 | } | |
560 | ||
397d425d EB |
561 | /** |
562 | * path_connected - Verify that a path->dentry is below path->mnt.mnt_root | |
563 | * @path: nameidate to verify | |
564 | * | |
565 | * Rename can sometimes move a file or directory outside of a bind | |
566 | * mount, path_connected allows those cases to be detected. | |
567 | */ | |
568 | static bool path_connected(const struct path *path) | |
569 | { | |
570 | struct vfsmount *mnt = path->mnt; | |
571 | ||
572 | /* Only bind mounts can have disconnected paths */ | |
573 | if (mnt->mnt_root == mnt->mnt_sb->s_root) | |
574 | return true; | |
575 | ||
576 | return is_subdir(path->dentry, mnt->mnt_root); | |
577 | } | |
578 | ||
894bc8c4 AV |
579 | static inline int nd_alloc_stack(struct nameidata *nd) |
580 | { | |
da4e0be0 | 581 | if (likely(nd->depth != EMBEDDED_LEVELS)) |
894bc8c4 AV |
582 | return 0; |
583 | if (likely(nd->stack != nd->internal)) | |
584 | return 0; | |
585 | return __nd_alloc_stack(nd); | |
586 | } | |
587 | ||
7973387a AV |
588 | static void drop_links(struct nameidata *nd) |
589 | { | |
590 | int i = nd->depth; | |
591 | while (i--) { | |
592 | struct saved *last = nd->stack + i; | |
fceef393 AV |
593 | do_delayed_call(&last->done); |
594 | clear_delayed_call(&last->done); | |
7973387a AV |
595 | } |
596 | } | |
597 | ||
598 | static void terminate_walk(struct nameidata *nd) | |
599 | { | |
600 | drop_links(nd); | |
601 | if (!(nd->flags & LOOKUP_RCU)) { | |
602 | int i; | |
603 | path_put(&nd->path); | |
604 | for (i = 0; i < nd->depth; i++) | |
605 | path_put(&nd->stack[i].link); | |
102b8af2 AV |
606 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
607 | path_put(&nd->root); | |
608 | nd->root.mnt = NULL; | |
609 | } | |
7973387a AV |
610 | } else { |
611 | nd->flags &= ~LOOKUP_RCU; | |
612 | if (!(nd->flags & LOOKUP_ROOT)) | |
613 | nd->root.mnt = NULL; | |
614 | rcu_read_unlock(); | |
615 | } | |
616 | nd->depth = 0; | |
617 | } | |
618 | ||
619 | /* path_put is needed afterwards regardless of success or failure */ | |
620 | static bool legitimize_path(struct nameidata *nd, | |
621 | struct path *path, unsigned seq) | |
622 | { | |
623 | int res = __legitimize_mnt(path->mnt, nd->m_seq); | |
624 | if (unlikely(res)) { | |
625 | if (res > 0) | |
626 | path->mnt = NULL; | |
627 | path->dentry = NULL; | |
628 | return false; | |
629 | } | |
630 | if (unlikely(!lockref_get_not_dead(&path->dentry->d_lockref))) { | |
631 | path->dentry = NULL; | |
632 | return false; | |
633 | } | |
634 | return !read_seqcount_retry(&path->dentry->d_seq, seq); | |
635 | } | |
636 | ||
637 | static bool legitimize_links(struct nameidata *nd) | |
638 | { | |
639 | int i; | |
640 | for (i = 0; i < nd->depth; i++) { | |
641 | struct saved *last = nd->stack + i; | |
642 | if (unlikely(!legitimize_path(nd, &last->link, last->seq))) { | |
643 | drop_links(nd); | |
644 | nd->depth = i + 1; | |
645 | return false; | |
646 | } | |
647 | } | |
648 | return true; | |
649 | } | |
650 | ||
19660af7 | 651 | /* |
31e6b01f | 652 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
653 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
654 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
57e3715c | 655 | * normal reference counts on dentries and vfsmounts to transition to ref-walk |
19660af7 AV |
656 | * mode. Refcounts are grabbed at the last known good point before rcu-walk |
657 | * got stuck, so ref-walk may continue from there. If this is not successful | |
658 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
659 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 660 | */ |
31e6b01f NP |
661 | |
662 | /** | |
19660af7 AV |
663 | * unlazy_walk - try to switch to ref-walk mode. |
664 | * @nd: nameidata pathwalk data | |
665 | * @dentry: child of nd->path.dentry or NULL | |
6e9918b7 | 666 | * @seq: seq number to check dentry against |
39191628 | 667 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 668 | * |
19660af7 AV |
669 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
670 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
671 | * @nd or NULL. Must be called from rcu-walk context. | |
7973387a AV |
672 | * Nothing should touch nameidata between unlazy_walk() failure and |
673 | * terminate_walk(). | |
31e6b01f | 674 | */ |
6e9918b7 | 675 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry, unsigned seq) |
31e6b01f | 676 | { |
31e6b01f NP |
677 | struct dentry *parent = nd->path.dentry; |
678 | ||
679 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 | 680 | |
e5c832d5 | 681 | nd->flags &= ~LOOKUP_RCU; |
7973387a AV |
682 | if (unlikely(!legitimize_links(nd))) |
683 | goto out2; | |
684 | if (unlikely(!legitimize_mnt(nd->path.mnt, nd->m_seq))) | |
685 | goto out2; | |
686 | if (unlikely(!lockref_get_not_dead(&parent->d_lockref))) | |
687 | goto out1; | |
48a066e7 | 688 | |
15570086 LT |
689 | /* |
690 | * For a negative lookup, the lookup sequence point is the parents | |
691 | * sequence point, and it only needs to revalidate the parent dentry. | |
692 | * | |
693 | * For a positive lookup, we need to move both the parent and the | |
694 | * dentry from the RCU domain to be properly refcounted. And the | |
695 | * sequence number in the dentry validates *both* dentry counters, | |
696 | * since we checked the sequence number of the parent after we got | |
697 | * the child sequence number. So we know the parent must still | |
698 | * be valid if the child sequence number is still valid. | |
699 | */ | |
19660af7 | 700 | if (!dentry) { |
e5c832d5 LT |
701 | if (read_seqcount_retry(&parent->d_seq, nd->seq)) |
702 | goto out; | |
19660af7 AV |
703 | BUG_ON(nd->inode != parent->d_inode); |
704 | } else { | |
e5c832d5 LT |
705 | if (!lockref_get_not_dead(&dentry->d_lockref)) |
706 | goto out; | |
6e9918b7 | 707 | if (read_seqcount_retry(&dentry->d_seq, seq)) |
e5c832d5 | 708 | goto drop_dentry; |
19660af7 | 709 | } |
e5c832d5 LT |
710 | |
711 | /* | |
712 | * Sequence counts matched. Now make sure that the root is | |
713 | * still valid and get it if required. | |
714 | */ | |
715 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
5a8d87e8 AV |
716 | if (unlikely(!legitimize_path(nd, &nd->root, nd->root_seq))) { |
717 | rcu_read_unlock(); | |
718 | dput(dentry); | |
719 | return -ECHILD; | |
7973387a | 720 | } |
31e6b01f | 721 | } |
31e6b01f | 722 | |
8b61e74f | 723 | rcu_read_unlock(); |
31e6b01f | 724 | return 0; |
19660af7 | 725 | |
e5c832d5 | 726 | drop_dentry: |
8b61e74f | 727 | rcu_read_unlock(); |
15570086 | 728 | dput(dentry); |
d0d27277 | 729 | goto drop_root_mnt; |
7973387a AV |
730 | out2: |
731 | nd->path.mnt = NULL; | |
732 | out1: | |
733 | nd->path.dentry = NULL; | |
e5c832d5 | 734 | out: |
8b61e74f | 735 | rcu_read_unlock(); |
d0d27277 LT |
736 | drop_root_mnt: |
737 | if (!(nd->flags & LOOKUP_ROOT)) | |
738 | nd->root.mnt = NULL; | |
31e6b01f NP |
739 | return -ECHILD; |
740 | } | |
741 | ||
7973387a AV |
742 | static int unlazy_link(struct nameidata *nd, struct path *link, unsigned seq) |
743 | { | |
744 | if (unlikely(!legitimize_path(nd, link, seq))) { | |
745 | drop_links(nd); | |
746 | nd->depth = 0; | |
747 | nd->flags &= ~LOOKUP_RCU; | |
748 | nd->path.mnt = NULL; | |
749 | nd->path.dentry = NULL; | |
750 | if (!(nd->flags & LOOKUP_ROOT)) | |
751 | nd->root.mnt = NULL; | |
752 | rcu_read_unlock(); | |
753 | } else if (likely(unlazy_walk(nd, NULL, 0)) == 0) { | |
754 | return 0; | |
755 | } | |
756 | path_put(link); | |
757 | return -ECHILD; | |
758 | } | |
759 | ||
4ce16ef3 | 760 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 761 | { |
4ce16ef3 | 762 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
763 | } |
764 | ||
9f1fafee AV |
765 | /** |
766 | * complete_walk - successful completion of path walk | |
767 | * @nd: pointer nameidata | |
39159de2 | 768 | * |
9f1fafee AV |
769 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
770 | * Revalidate the final result, unless we'd already done that during | |
771 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
772 | * success, -error on failure. In case of failure caller does not | |
773 | * need to drop nd->path. | |
39159de2 | 774 | */ |
9f1fafee | 775 | static int complete_walk(struct nameidata *nd) |
39159de2 | 776 | { |
16c2cd71 | 777 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 778 | int status; |
39159de2 | 779 | |
9f1fafee | 780 | if (nd->flags & LOOKUP_RCU) { |
9f1fafee AV |
781 | if (!(nd->flags & LOOKUP_ROOT)) |
782 | nd->root.mnt = NULL; | |
6e9918b7 | 783 | if (unlikely(unlazy_walk(nd, NULL, 0))) |
9f1fafee | 784 | return -ECHILD; |
9f1fafee AV |
785 | } |
786 | ||
16c2cd71 AV |
787 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
788 | return 0; | |
789 | ||
ecf3d1f1 | 790 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
791 | return 0; |
792 | ||
ecf3d1f1 | 793 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
794 | if (status > 0) |
795 | return 0; | |
796 | ||
16c2cd71 | 797 | if (!status) |
39159de2 | 798 | status = -ESTALE; |
16c2cd71 | 799 | |
39159de2 JL |
800 | return status; |
801 | } | |
802 | ||
18d8c860 | 803 | static void set_root(struct nameidata *nd) |
31e6b01f | 804 | { |
7bd88377 | 805 | struct fs_struct *fs = current->fs; |
c28cc364 | 806 | |
9e6697e2 AV |
807 | if (nd->flags & LOOKUP_RCU) { |
808 | unsigned seq; | |
809 | ||
810 | do { | |
811 | seq = read_seqcount_begin(&fs->seq); | |
812 | nd->root = fs->root; | |
813 | nd->root_seq = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
814 | } while (read_seqcount_retry(&fs->seq, seq)); | |
815 | } else { | |
816 | get_fs_root(fs, &nd->root); | |
817 | } | |
31e6b01f NP |
818 | } |
819 | ||
1d957f9b | 820 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
821 | { |
822 | dput(path->dentry); | |
4ac91378 | 823 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
824 | mntput(path->mnt); |
825 | } | |
826 | ||
7b9337aa NP |
827 | static inline void path_to_nameidata(const struct path *path, |
828 | struct nameidata *nd) | |
051d3812 | 829 | { |
31e6b01f NP |
830 | if (!(nd->flags & LOOKUP_RCU)) { |
831 | dput(nd->path.dentry); | |
832 | if (nd->path.mnt != path->mnt) | |
833 | mntput(nd->path.mnt); | |
9a229683 | 834 | } |
31e6b01f | 835 | nd->path.mnt = path->mnt; |
4ac91378 | 836 | nd->path.dentry = path->dentry; |
051d3812 IK |
837 | } |
838 | ||
248fb5b9 AV |
839 | static int nd_jump_root(struct nameidata *nd) |
840 | { | |
841 | if (nd->flags & LOOKUP_RCU) { | |
842 | struct dentry *d; | |
843 | nd->path = nd->root; | |
844 | d = nd->path.dentry; | |
845 | nd->inode = d->d_inode; | |
846 | nd->seq = nd->root_seq; | |
847 | if (unlikely(read_seqcount_retry(&d->d_seq, nd->seq))) | |
848 | return -ECHILD; | |
849 | } else { | |
850 | path_put(&nd->path); | |
851 | nd->path = nd->root; | |
852 | path_get(&nd->path); | |
853 | nd->inode = nd->path.dentry->d_inode; | |
854 | } | |
855 | nd->flags |= LOOKUP_JUMPED; | |
856 | return 0; | |
857 | } | |
858 | ||
b5fb63c1 | 859 | /* |
6b255391 | 860 | * Helper to directly jump to a known parsed path from ->get_link, |
b5fb63c1 CH |
861 | * caller must have taken a reference to path beforehand. |
862 | */ | |
6e77137b | 863 | void nd_jump_link(struct path *path) |
b5fb63c1 | 864 | { |
6e77137b | 865 | struct nameidata *nd = current->nameidata; |
b5fb63c1 CH |
866 | path_put(&nd->path); |
867 | ||
868 | nd->path = *path; | |
869 | nd->inode = nd->path.dentry->d_inode; | |
870 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
871 | } |
872 | ||
b9ff4429 | 873 | static inline void put_link(struct nameidata *nd) |
574197e0 | 874 | { |
21c3003d | 875 | struct saved *last = nd->stack + --nd->depth; |
fceef393 | 876 | do_delayed_call(&last->done); |
6548fae2 AV |
877 | if (!(nd->flags & LOOKUP_RCU)) |
878 | path_put(&last->link); | |
574197e0 AV |
879 | } |
880 | ||
561ec64a LT |
881 | int sysctl_protected_symlinks __read_mostly = 0; |
882 | int sysctl_protected_hardlinks __read_mostly = 0; | |
800179c9 KC |
883 | |
884 | /** | |
885 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 886 | * @nd: nameidata pathwalk data |
800179c9 KC |
887 | * |
888 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
889 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
890 | * in a sticky world-writable directory. This is to protect privileged | |
891 | * processes from failing races against path names that may change out | |
892 | * from under them by way of other users creating malicious symlinks. | |
893 | * It will permit symlinks to be followed only when outside a sticky | |
894 | * world-writable directory, or when the uid of the symlink and follower | |
895 | * match, or when the directory owner matches the symlink's owner. | |
896 | * | |
897 | * Returns 0 if following the symlink is allowed, -ve on error. | |
898 | */ | |
fec2fa24 | 899 | static inline int may_follow_link(struct nameidata *nd) |
800179c9 KC |
900 | { |
901 | const struct inode *inode; | |
902 | const struct inode *parent; | |
903 | ||
904 | if (!sysctl_protected_symlinks) | |
905 | return 0; | |
906 | ||
907 | /* Allowed if owner and follower match. */ | |
fceef393 | 908 | inode = nd->link_inode; |
81abe27b | 909 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
910 | return 0; |
911 | ||
912 | /* Allowed if parent directory not sticky and world-writable. */ | |
aa65fa35 | 913 | parent = nd->inode; |
800179c9 KC |
914 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) |
915 | return 0; | |
916 | ||
917 | /* Allowed if parent directory and link owner match. */ | |
81abe27b | 918 | if (uid_eq(parent->i_uid, inode->i_uid)) |
800179c9 KC |
919 | return 0; |
920 | ||
31956502 AV |
921 | if (nd->flags & LOOKUP_RCU) |
922 | return -ECHILD; | |
923 | ||
1cf2665b | 924 | audit_log_link_denied("follow_link", &nd->stack[0].link); |
800179c9 KC |
925 | return -EACCES; |
926 | } | |
927 | ||
928 | /** | |
929 | * safe_hardlink_source - Check for safe hardlink conditions | |
930 | * @inode: the source inode to hardlink from | |
931 | * | |
932 | * Return false if at least one of the following conditions: | |
933 | * - inode is not a regular file | |
934 | * - inode is setuid | |
935 | * - inode is setgid and group-exec | |
936 | * - access failure for read and write | |
937 | * | |
938 | * Otherwise returns true. | |
939 | */ | |
940 | static bool safe_hardlink_source(struct inode *inode) | |
941 | { | |
942 | umode_t mode = inode->i_mode; | |
943 | ||
944 | /* Special files should not get pinned to the filesystem. */ | |
945 | if (!S_ISREG(mode)) | |
946 | return false; | |
947 | ||
948 | /* Setuid files should not get pinned to the filesystem. */ | |
949 | if (mode & S_ISUID) | |
950 | return false; | |
951 | ||
952 | /* Executable setgid files should not get pinned to the filesystem. */ | |
953 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
954 | return false; | |
955 | ||
956 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
957 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
958 | return false; | |
959 | ||
960 | return true; | |
961 | } | |
962 | ||
963 | /** | |
964 | * may_linkat - Check permissions for creating a hardlink | |
965 | * @link: the source to hardlink from | |
966 | * | |
967 | * Block hardlink when all of: | |
968 | * - sysctl_protected_hardlinks enabled | |
969 | * - fsuid does not match inode | |
970 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
f2ca3796 | 971 | * - not CAP_FOWNER in a namespace with the inode owner uid mapped |
800179c9 KC |
972 | * |
973 | * Returns 0 if successful, -ve on error. | |
974 | */ | |
975 | static int may_linkat(struct path *link) | |
976 | { | |
800179c9 KC |
977 | struct inode *inode; |
978 | ||
979 | if (!sysctl_protected_hardlinks) | |
980 | return 0; | |
981 | ||
800179c9 KC |
982 | inode = link->dentry->d_inode; |
983 | ||
984 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
985 | * otherwise, it must be a safe source. | |
986 | */ | |
f2ca3796 | 987 | if (inode_owner_or_capable(inode) || safe_hardlink_source(inode)) |
800179c9 KC |
988 | return 0; |
989 | ||
a51d9eaa | 990 | audit_log_link_denied("linkat", link); |
800179c9 KC |
991 | return -EPERM; |
992 | } | |
993 | ||
3b2e7f75 AV |
994 | static __always_inline |
995 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 996 | { |
ab104923 | 997 | struct saved *last = nd->stack + nd->depth - 1; |
1cf2665b | 998 | struct dentry *dentry = last->link.dentry; |
fceef393 | 999 | struct inode *inode = nd->link_inode; |
6d7b5aae | 1000 | int error; |
0a959df5 | 1001 | const char *res; |
1da177e4 | 1002 | |
8fa9dd24 N |
1003 | if (!(nd->flags & LOOKUP_RCU)) { |
1004 | touch_atime(&last->link); | |
1005 | cond_resched(); | |
1006 | } else if (atime_needs_update(&last->link, inode)) { | |
bc40aee0 AV |
1007 | if (unlikely(unlazy_walk(nd, NULL, 0))) |
1008 | return ERR_PTR(-ECHILD); | |
8fa9dd24 | 1009 | touch_atime(&last->link); |
bc40aee0 | 1010 | } |
574197e0 | 1011 | |
bda0be7a N |
1012 | error = security_inode_follow_link(dentry, inode, |
1013 | nd->flags & LOOKUP_RCU); | |
1014 | if (unlikely(error)) | |
6920a440 | 1015 | return ERR_PTR(error); |
36f3b4f6 | 1016 | |
86acdca1 | 1017 | nd->last_type = LAST_BIND; |
d4dee48b AV |
1018 | res = inode->i_link; |
1019 | if (!res) { | |
fceef393 AV |
1020 | const char * (*get)(struct dentry *, struct inode *, |
1021 | struct delayed_call *); | |
1022 | get = inode->i_op->get_link; | |
8c1b4566 | 1023 | if (nd->flags & LOOKUP_RCU) { |
fceef393 | 1024 | res = get(NULL, inode, &last->done); |
6b255391 AV |
1025 | if (res == ERR_PTR(-ECHILD)) { |
1026 | if (unlikely(unlazy_walk(nd, NULL, 0))) | |
1027 | return ERR_PTR(-ECHILD); | |
fceef393 | 1028 | res = get(dentry, inode, &last->done); |
6b255391 AV |
1029 | } |
1030 | } else { | |
fceef393 | 1031 | res = get(dentry, inode, &last->done); |
8c1b4566 | 1032 | } |
fceef393 | 1033 | if (IS_ERR_OR_NULL(res)) |
fab51e8a | 1034 | return res; |
fab51e8a AV |
1035 | } |
1036 | if (*res == '/') { | |
9e6697e2 AV |
1037 | if (!nd->root.mnt) |
1038 | set_root(nd); | |
248fb5b9 AV |
1039 | if (unlikely(nd_jump_root(nd))) |
1040 | return ERR_PTR(-ECHILD); | |
fab51e8a AV |
1041 | while (unlikely(*++res == '/')) |
1042 | ; | |
1da177e4 | 1043 | } |
fab51e8a AV |
1044 | if (!*res) |
1045 | res = NULL; | |
0a959df5 AV |
1046 | return res; |
1047 | } | |
6d7b5aae | 1048 | |
f015f126 DH |
1049 | /* |
1050 | * follow_up - Find the mountpoint of path's vfsmount | |
1051 | * | |
1052 | * Given a path, find the mountpoint of its source file system. | |
1053 | * Replace @path with the path of the mountpoint in the parent mount. | |
1054 | * Up is towards /. | |
1055 | * | |
1056 | * Return 1 if we went up a level and 0 if we were already at the | |
1057 | * root. | |
1058 | */ | |
bab77ebf | 1059 | int follow_up(struct path *path) |
1da177e4 | 1060 | { |
0714a533 AV |
1061 | struct mount *mnt = real_mount(path->mnt); |
1062 | struct mount *parent; | |
1da177e4 | 1063 | struct dentry *mountpoint; |
99b7db7b | 1064 | |
48a066e7 | 1065 | read_seqlock_excl(&mount_lock); |
0714a533 | 1066 | parent = mnt->mnt_parent; |
3c0a6163 | 1067 | if (parent == mnt) { |
48a066e7 | 1068 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
1069 | return 0; |
1070 | } | |
0714a533 | 1071 | mntget(&parent->mnt); |
a73324da | 1072 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 1073 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
1074 | dput(path->dentry); |
1075 | path->dentry = mountpoint; | |
1076 | mntput(path->mnt); | |
0714a533 | 1077 | path->mnt = &parent->mnt; |
1da177e4 LT |
1078 | return 1; |
1079 | } | |
4d359507 | 1080 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 1081 | |
b5c84bf6 | 1082 | /* |
9875cf80 DH |
1083 | * Perform an automount |
1084 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
1085 | * were called with. | |
1da177e4 | 1086 | */ |
756daf26 | 1087 | static int follow_automount(struct path *path, struct nameidata *nd, |
9875cf80 | 1088 | bool *need_mntput) |
31e6b01f | 1089 | { |
9875cf80 | 1090 | struct vfsmount *mnt; |
ea5b778a | 1091 | int err; |
9875cf80 DH |
1092 | |
1093 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
1094 | return -EREMOTE; | |
1095 | ||
0ec26fd0 MS |
1096 | /* We don't want to mount if someone's just doing a stat - |
1097 | * unless they're stat'ing a directory and appended a '/' to | |
1098 | * the name. | |
1099 | * | |
1100 | * We do, however, want to mount if someone wants to open or | |
1101 | * create a file of any type under the mountpoint, wants to | |
1102 | * traverse through the mountpoint or wants to open the | |
1103 | * mounted directory. Also, autofs may mark negative dentries | |
1104 | * as being automount points. These will need the attentions | |
1105 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 1106 | */ |
756daf26 N |
1107 | if (!(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
1108 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && | |
0ec26fd0 MS |
1109 | path->dentry->d_inode) |
1110 | return -EISDIR; | |
1111 | ||
756daf26 N |
1112 | nd->total_link_count++; |
1113 | if (nd->total_link_count >= 40) | |
9875cf80 DH |
1114 | return -ELOOP; |
1115 | ||
1116 | mnt = path->dentry->d_op->d_automount(path); | |
1117 | if (IS_ERR(mnt)) { | |
1118 | /* | |
1119 | * The filesystem is allowed to return -EISDIR here to indicate | |
1120 | * it doesn't want to automount. For instance, autofs would do | |
1121 | * this so that its userspace daemon can mount on this dentry. | |
1122 | * | |
1123 | * However, we can only permit this if it's a terminal point in | |
1124 | * the path being looked up; if it wasn't then the remainder of | |
1125 | * the path is inaccessible and we should say so. | |
1126 | */ | |
756daf26 | 1127 | if (PTR_ERR(mnt) == -EISDIR && (nd->flags & LOOKUP_PARENT)) |
9875cf80 DH |
1128 | return -EREMOTE; |
1129 | return PTR_ERR(mnt); | |
31e6b01f | 1130 | } |
ea5b778a | 1131 | |
9875cf80 DH |
1132 | if (!mnt) /* mount collision */ |
1133 | return 0; | |
31e6b01f | 1134 | |
8aef1884 AV |
1135 | if (!*need_mntput) { |
1136 | /* lock_mount() may release path->mnt on error */ | |
1137 | mntget(path->mnt); | |
1138 | *need_mntput = true; | |
1139 | } | |
19a167af | 1140 | err = finish_automount(mnt, path); |
9875cf80 | 1141 | |
ea5b778a DH |
1142 | switch (err) { |
1143 | case -EBUSY: | |
1144 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1145 | return 0; |
ea5b778a | 1146 | case 0: |
8aef1884 | 1147 | path_put(path); |
ea5b778a DH |
1148 | path->mnt = mnt; |
1149 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1150 | return 0; |
19a167af AV |
1151 | default: |
1152 | return err; | |
ea5b778a | 1153 | } |
19a167af | 1154 | |
463ffb2e AV |
1155 | } |
1156 | ||
9875cf80 DH |
1157 | /* |
1158 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1159 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1160 | * - Flagged as mountpoint |
1161 | * - Flagged as automount point | |
1162 | * | |
1163 | * This may only be called in refwalk mode. | |
1164 | * | |
1165 | * Serialization is taken care of in namespace.c | |
1166 | */ | |
756daf26 | 1167 | static int follow_managed(struct path *path, struct nameidata *nd) |
1da177e4 | 1168 | { |
8aef1884 | 1169 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1170 | unsigned managed; |
1171 | bool need_mntput = false; | |
8aef1884 | 1172 | int ret = 0; |
9875cf80 DH |
1173 | |
1174 | /* Given that we're not holding a lock here, we retain the value in a | |
1175 | * local variable for each dentry as we look at it so that we don't see | |
1176 | * the components of that value change under us */ | |
1177 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1178 | managed &= DCACHE_MANAGED_DENTRY, | |
1179 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1180 | /* Allow the filesystem to manage the transit without i_mutex |
1181 | * being held. */ | |
1182 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1183 | BUG_ON(!path->dentry->d_op); | |
1184 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 1185 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 1186 | if (ret < 0) |
8aef1884 | 1187 | break; |
cc53ce53 DH |
1188 | } |
1189 | ||
9875cf80 DH |
1190 | /* Transit to a mounted filesystem. */ |
1191 | if (managed & DCACHE_MOUNTED) { | |
1192 | struct vfsmount *mounted = lookup_mnt(path); | |
1193 | if (mounted) { | |
1194 | dput(path->dentry); | |
1195 | if (need_mntput) | |
1196 | mntput(path->mnt); | |
1197 | path->mnt = mounted; | |
1198 | path->dentry = dget(mounted->mnt_root); | |
1199 | need_mntput = true; | |
1200 | continue; | |
1201 | } | |
1202 | ||
1203 | /* Something is mounted on this dentry in another | |
1204 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1205 | * namespace got unmounted before lookup_mnt() could |
1206 | * get it */ | |
9875cf80 DH |
1207 | } |
1208 | ||
1209 | /* Handle an automount point */ | |
1210 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
756daf26 | 1211 | ret = follow_automount(path, nd, &need_mntput); |
9875cf80 | 1212 | if (ret < 0) |
8aef1884 | 1213 | break; |
9875cf80 DH |
1214 | continue; |
1215 | } | |
1216 | ||
1217 | /* We didn't change the current path point */ | |
1218 | break; | |
1da177e4 | 1219 | } |
8aef1884 AV |
1220 | |
1221 | if (need_mntput && path->mnt == mnt) | |
1222 | mntput(path->mnt); | |
e9742b53 AV |
1223 | if (ret == -EISDIR || !ret) |
1224 | ret = 1; | |
8402752e AV |
1225 | if (need_mntput) |
1226 | nd->flags |= LOOKUP_JUMPED; | |
1227 | if (unlikely(ret < 0)) | |
1228 | path_put_conditional(path, nd); | |
1229 | return ret; | |
1da177e4 LT |
1230 | } |
1231 | ||
cc53ce53 | 1232 | int follow_down_one(struct path *path) |
1da177e4 LT |
1233 | { |
1234 | struct vfsmount *mounted; | |
1235 | ||
1c755af4 | 1236 | mounted = lookup_mnt(path); |
1da177e4 | 1237 | if (mounted) { |
9393bd07 AV |
1238 | dput(path->dentry); |
1239 | mntput(path->mnt); | |
1240 | path->mnt = mounted; | |
1241 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1242 | return 1; |
1243 | } | |
1244 | return 0; | |
1245 | } | |
4d359507 | 1246 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1247 | |
b8faf035 | 1248 | static inline int managed_dentry_rcu(struct dentry *dentry) |
62a7375e | 1249 | { |
b8faf035 N |
1250 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1251 | dentry->d_op->d_manage(dentry, true) : 0; | |
62a7375e IK |
1252 | } |
1253 | ||
9875cf80 | 1254 | /* |
287548e4 AV |
1255 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1256 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1257 | */ |
1258 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1259 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1260 | { |
62a7375e | 1261 | for (;;) { |
c7105365 | 1262 | struct mount *mounted; |
62a7375e IK |
1263 | /* |
1264 | * Don't forget we might have a non-mountpoint managed dentry | |
1265 | * that wants to block transit. | |
1266 | */ | |
b8faf035 N |
1267 | switch (managed_dentry_rcu(path->dentry)) { |
1268 | case -ECHILD: | |
1269 | default: | |
ab90911f | 1270 | return false; |
b8faf035 N |
1271 | case -EISDIR: |
1272 | return true; | |
1273 | case 0: | |
1274 | break; | |
1275 | } | |
62a7375e IK |
1276 | |
1277 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1278 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1279 | |
474279dc | 1280 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1281 | if (!mounted) |
1282 | break; | |
c7105365 AV |
1283 | path->mnt = &mounted->mnt; |
1284 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1285 | nd->flags |= LOOKUP_JUMPED; |
254cf582 | 1286 | *seqp = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1287 | /* |
1288 | * Update the inode too. We don't need to re-check the | |
1289 | * dentry sequence number here after this d_inode read, | |
1290 | * because a mount-point is always pinned. | |
1291 | */ | |
1292 | *inode = path->dentry->d_inode; | |
9875cf80 | 1293 | } |
f5be3e29 | 1294 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1295 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1296 | } |
1297 | ||
31e6b01f NP |
1298 | static int follow_dotdot_rcu(struct nameidata *nd) |
1299 | { | |
4023bfc9 | 1300 | struct inode *inode = nd->inode; |
31e6b01f | 1301 | |
9875cf80 | 1302 | while (1) { |
aed434ad | 1303 | if (path_equal(&nd->path, &nd->root)) |
31e6b01f | 1304 | break; |
31e6b01f NP |
1305 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
1306 | struct dentry *old = nd->path.dentry; | |
1307 | struct dentry *parent = old->d_parent; | |
1308 | unsigned seq; | |
1309 | ||
4023bfc9 | 1310 | inode = parent->d_inode; |
31e6b01f | 1311 | seq = read_seqcount_begin(&parent->d_seq); |
aed434ad AV |
1312 | if (unlikely(read_seqcount_retry(&old->d_seq, nd->seq))) |
1313 | return -ECHILD; | |
31e6b01f NP |
1314 | nd->path.dentry = parent; |
1315 | nd->seq = seq; | |
397d425d EB |
1316 | if (unlikely(!path_connected(&nd->path))) |
1317 | return -ENOENT; | |
31e6b01f | 1318 | break; |
aed434ad AV |
1319 | } else { |
1320 | struct mount *mnt = real_mount(nd->path.mnt); | |
1321 | struct mount *mparent = mnt->mnt_parent; | |
1322 | struct dentry *mountpoint = mnt->mnt_mountpoint; | |
1323 | struct inode *inode2 = mountpoint->d_inode; | |
1324 | unsigned seq = read_seqcount_begin(&mountpoint->d_seq); | |
1325 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) | |
1326 | return -ECHILD; | |
1327 | if (&mparent->mnt == nd->path.mnt) | |
1328 | break; | |
1329 | /* we know that mountpoint was pinned */ | |
1330 | nd->path.dentry = mountpoint; | |
1331 | nd->path.mnt = &mparent->mnt; | |
1332 | inode = inode2; | |
1333 | nd->seq = seq; | |
31e6b01f | 1334 | } |
31e6b01f | 1335 | } |
aed434ad | 1336 | while (unlikely(d_mountpoint(nd->path.dentry))) { |
b37199e6 AV |
1337 | struct mount *mounted; |
1338 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
aed434ad AV |
1339 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) |
1340 | return -ECHILD; | |
b37199e6 AV |
1341 | if (!mounted) |
1342 | break; | |
1343 | nd->path.mnt = &mounted->mnt; | |
1344 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1345 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1346 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
b37199e6 | 1347 | } |
4023bfc9 | 1348 | nd->inode = inode; |
31e6b01f NP |
1349 | return 0; |
1350 | } | |
1351 | ||
cc53ce53 DH |
1352 | /* |
1353 | * Follow down to the covering mount currently visible to userspace. At each | |
1354 | * point, the filesystem owning that dentry may be queried as to whether the | |
1355 | * caller is permitted to proceed or not. | |
cc53ce53 | 1356 | */ |
7cc90cc3 | 1357 | int follow_down(struct path *path) |
cc53ce53 DH |
1358 | { |
1359 | unsigned managed; | |
1360 | int ret; | |
1361 | ||
1362 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1363 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1364 | /* Allow the filesystem to manage the transit without i_mutex | |
1365 | * being held. | |
1366 | * | |
1367 | * We indicate to the filesystem if someone is trying to mount | |
1368 | * something here. This gives autofs the chance to deny anyone | |
1369 | * other than its daemon the right to mount on its | |
1370 | * superstructure. | |
1371 | * | |
1372 | * The filesystem may sleep at this point. | |
1373 | */ | |
1374 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1375 | BUG_ON(!path->dentry->d_op); | |
1376 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1377 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1378 | path->dentry, false); |
cc53ce53 DH |
1379 | if (ret < 0) |
1380 | return ret == -EISDIR ? 0 : ret; | |
1381 | } | |
1382 | ||
1383 | /* Transit to a mounted filesystem. */ | |
1384 | if (managed & DCACHE_MOUNTED) { | |
1385 | struct vfsmount *mounted = lookup_mnt(path); | |
1386 | if (!mounted) | |
1387 | break; | |
1388 | dput(path->dentry); | |
1389 | mntput(path->mnt); | |
1390 | path->mnt = mounted; | |
1391 | path->dentry = dget(mounted->mnt_root); | |
1392 | continue; | |
1393 | } | |
1394 | ||
1395 | /* Don't handle automount points here */ | |
1396 | break; | |
1397 | } | |
1398 | return 0; | |
1399 | } | |
4d359507 | 1400 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1401 | |
9875cf80 DH |
1402 | /* |
1403 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1404 | */ | |
1405 | static void follow_mount(struct path *path) | |
1406 | { | |
1407 | while (d_mountpoint(path->dentry)) { | |
1408 | struct vfsmount *mounted = lookup_mnt(path); | |
1409 | if (!mounted) | |
1410 | break; | |
1411 | dput(path->dentry); | |
1412 | mntput(path->mnt); | |
1413 | path->mnt = mounted; | |
1414 | path->dentry = dget(mounted->mnt_root); | |
1415 | } | |
1416 | } | |
1417 | ||
397d425d | 1418 | static int follow_dotdot(struct nameidata *nd) |
1da177e4 LT |
1419 | { |
1420 | while(1) { | |
4ac91378 | 1421 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1422 | |
2a737871 AV |
1423 | if (nd->path.dentry == nd->root.dentry && |
1424 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1425 | break; |
1426 | } | |
4ac91378 | 1427 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1428 | /* rare case of legitimate dget_parent()... */ |
1429 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 | 1430 | dput(old); |
397d425d EB |
1431 | if (unlikely(!path_connected(&nd->path))) |
1432 | return -ENOENT; | |
1da177e4 LT |
1433 | break; |
1434 | } | |
3088dd70 | 1435 | if (!follow_up(&nd->path)) |
1da177e4 | 1436 | break; |
1da177e4 | 1437 | } |
79ed0226 | 1438 | follow_mount(&nd->path); |
31e6b01f | 1439 | nd->inode = nd->path.dentry->d_inode; |
397d425d | 1440 | return 0; |
1da177e4 LT |
1441 | } |
1442 | ||
baa03890 | 1443 | /* |
bad61189 MS |
1444 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1445 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1446 | * returns whether i_op->lookup is necessary. | |
1447 | * | |
1448 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1449 | */ |
bad61189 | 1450 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
6c51e513 | 1451 | unsigned int flags) |
baa03890 | 1452 | { |
baa03890 | 1453 | struct dentry *dentry; |
bad61189 | 1454 | int error; |
baa03890 | 1455 | |
bad61189 MS |
1456 | dentry = d_lookup(dir, name); |
1457 | if (dentry) { | |
39e3c955 | 1458 | if (dentry->d_flags & DCACHE_OP_REVALIDATE) { |
201f956e | 1459 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1460 | if (unlikely(error <= 0)) { |
1461 | if (error < 0) { | |
1462 | dput(dentry); | |
1463 | return ERR_PTR(error); | |
5542aa2f EB |
1464 | } else { |
1465 | d_invalidate(dentry); | |
bad61189 MS |
1466 | dput(dentry); |
1467 | dentry = NULL; | |
1468 | } | |
1469 | } | |
1470 | } | |
1471 | } | |
baa03890 NP |
1472 | return dentry; |
1473 | } | |
1474 | ||
44396f4b | 1475 | /* |
13a2c3be BF |
1476 | * Call i_op->lookup on the dentry. The dentry must be negative and |
1477 | * unhashed. | |
bad61189 MS |
1478 | * |
1479 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1480 | */ |
bad61189 | 1481 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1482 | unsigned int flags) |
44396f4b | 1483 | { |
44396f4b JB |
1484 | struct dentry *old; |
1485 | ||
1486 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1487 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1488 | dput(dentry); |
44396f4b | 1489 | return ERR_PTR(-ENOENT); |
e188dc02 | 1490 | } |
44396f4b | 1491 | |
72bd866a | 1492 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1493 | if (unlikely(old)) { |
1494 | dput(dentry); | |
1495 | dentry = old; | |
1496 | } | |
1497 | return dentry; | |
1498 | } | |
1499 | ||
a3255546 | 1500 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1501 | struct dentry *base, unsigned int flags) |
a3255546 | 1502 | { |
6c51e513 | 1503 | struct dentry *dentry = lookup_dcache(name, base, flags); |
a3255546 | 1504 | |
6c51e513 | 1505 | if (dentry) |
bad61189 | 1506 | return dentry; |
a3255546 | 1507 | |
6c51e513 AV |
1508 | dentry = d_alloc(base, name); |
1509 | if (unlikely(!dentry)) | |
1510 | return ERR_PTR(-ENOMEM); | |
1511 | ||
72bd866a | 1512 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1513 | } |
1514 | ||
e97cdc87 | 1515 | static int lookup_fast(struct nameidata *nd, |
254cf582 AV |
1516 | struct path *path, struct inode **inode, |
1517 | unsigned *seqp) | |
1da177e4 | 1518 | { |
4ac91378 | 1519 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1520 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 | 1521 | int status = 1; |
9875cf80 DH |
1522 | int err; |
1523 | ||
b04f784e NP |
1524 | /* |
1525 | * Rename seqlock is not required here because in the off chance | |
5d0f49c1 AV |
1526 | * of a false negative due to a concurrent rename, the caller is |
1527 | * going to fall back to non-racy lookup. | |
b04f784e | 1528 | */ |
31e6b01f NP |
1529 | if (nd->flags & LOOKUP_RCU) { |
1530 | unsigned seq; | |
766c4cbf | 1531 | bool negative; |
da53be12 | 1532 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5d0f49c1 AV |
1533 | if (unlikely(!dentry)) { |
1534 | if (unlazy_walk(nd, NULL, 0)) | |
1535 | return -ECHILD; | |
e9742b53 | 1536 | return 0; |
5d0f49c1 | 1537 | } |
5a18fff2 | 1538 | |
12f8ad4b LT |
1539 | /* |
1540 | * This sequence count validates that the inode matches | |
1541 | * the dentry name information from lookup. | |
1542 | */ | |
63afdfc7 | 1543 | *inode = d_backing_inode(dentry); |
766c4cbf | 1544 | negative = d_is_negative(dentry); |
5d0f49c1 | 1545 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
12f8ad4b LT |
1546 | return -ECHILD; |
1547 | ||
1548 | /* | |
1549 | * This sequence count validates that the parent had no | |
1550 | * changes while we did the lookup of the dentry above. | |
1551 | * | |
1552 | * The memory barrier in read_seqcount_begin of child is | |
1553 | * enough, we can use __read_seqcount_retry here. | |
1554 | */ | |
5d0f49c1 | 1555 | if (unlikely(__read_seqcount_retry(&parent->d_seq, nd->seq))) |
31e6b01f | 1556 | return -ECHILD; |
5a18fff2 | 1557 | |
254cf582 | 1558 | *seqp = seq; |
5d0f49c1 | 1559 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) |
4ce16ef3 | 1560 | status = d_revalidate(dentry, nd->flags); |
5d0f49c1 AV |
1561 | if (unlikely(status <= 0)) { |
1562 | if (unlazy_walk(nd, dentry, seq)) | |
1563 | return -ECHILD; | |
1564 | if (status == -ECHILD) | |
1565 | status = d_revalidate(dentry, nd->flags); | |
1566 | } else { | |
1567 | /* | |
1568 | * Note: do negative dentry check after revalidation in | |
1569 | * case that drops it. | |
1570 | */ | |
1571 | if (unlikely(negative)) | |
1572 | return -ENOENT; | |
1573 | path->mnt = mnt; | |
1574 | path->dentry = dentry; | |
1575 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) | |
e9742b53 | 1576 | return 1; |
5d0f49c1 AV |
1577 | if (unlazy_walk(nd, dentry, seq)) |
1578 | return -ECHILD; | |
24643087 | 1579 | } |
5a18fff2 | 1580 | } else { |
e97cdc87 | 1581 | dentry = __d_lookup(parent, &nd->last); |
5d0f49c1 | 1582 | if (unlikely(!dentry)) |
e9742b53 | 1583 | return 0; |
5d0f49c1 AV |
1584 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) |
1585 | status = d_revalidate(dentry, nd->flags); | |
9875cf80 | 1586 | } |
5a18fff2 | 1587 | if (unlikely(status <= 0)) { |
e9742b53 | 1588 | if (!status) |
5d0f49c1 | 1589 | d_invalidate(dentry); |
5542aa2f | 1590 | dput(dentry); |
5d0f49c1 | 1591 | return status; |
24643087 | 1592 | } |
766c4cbf AV |
1593 | if (unlikely(d_is_negative(dentry))) { |
1594 | dput(dentry); | |
1595 | return -ENOENT; | |
1596 | } | |
5d0f49c1 | 1597 | |
9875cf80 DH |
1598 | path->mnt = mnt; |
1599 | path->dentry = dentry; | |
756daf26 | 1600 | err = follow_managed(path, nd); |
e9742b53 | 1601 | if (likely(err > 0)) |
63afdfc7 | 1602 | *inode = d_backing_inode(path->dentry); |
8402752e | 1603 | return err; |
697f514d MS |
1604 | } |
1605 | ||
1606 | /* Fast lookup failed, do it the slow way */ | |
cc2a5271 | 1607 | static int lookup_slow(struct nameidata *nd, struct path *path) |
697f514d MS |
1608 | { |
1609 | struct dentry *dentry, *parent; | |
697f514d MS |
1610 | |
1611 | parent = nd->path.dentry; | |
81e6f520 AV |
1612 | BUG_ON(nd->inode != parent->d_inode); |
1613 | ||
5955102c | 1614 | inode_lock(parent->d_inode); |
cc2a5271 | 1615 | dentry = __lookup_hash(&nd->last, parent, nd->flags); |
5955102c | 1616 | inode_unlock(parent->d_inode); |
81e6f520 AV |
1617 | if (IS_ERR(dentry)) |
1618 | return PTR_ERR(dentry); | |
697f514d MS |
1619 | path->mnt = nd->path.mnt; |
1620 | path->dentry = dentry; | |
8402752e | 1621 | return follow_managed(path, nd); |
1da177e4 LT |
1622 | } |
1623 | ||
52094c8a AV |
1624 | static inline int may_lookup(struct nameidata *nd) |
1625 | { | |
1626 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1627 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1628 | if (err != -ECHILD) |
1629 | return err; | |
6e9918b7 | 1630 | if (unlazy_walk(nd, NULL, 0)) |
52094c8a AV |
1631 | return -ECHILD; |
1632 | } | |
4ad5abb3 | 1633 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1634 | } |
1635 | ||
9856fa1b AV |
1636 | static inline int handle_dots(struct nameidata *nd, int type) |
1637 | { | |
1638 | if (type == LAST_DOTDOT) { | |
9e6697e2 AV |
1639 | if (!nd->root.mnt) |
1640 | set_root(nd); | |
9856fa1b | 1641 | if (nd->flags & LOOKUP_RCU) { |
70291aec | 1642 | return follow_dotdot_rcu(nd); |
9856fa1b | 1643 | } else |
397d425d | 1644 | return follow_dotdot(nd); |
9856fa1b AV |
1645 | } |
1646 | return 0; | |
1647 | } | |
1648 | ||
181548c0 AV |
1649 | static int pick_link(struct nameidata *nd, struct path *link, |
1650 | struct inode *inode, unsigned seq) | |
d63ff28f | 1651 | { |
626de996 | 1652 | int error; |
1cf2665b | 1653 | struct saved *last; |
756daf26 | 1654 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) { |
626de996 AV |
1655 | path_to_nameidata(link, nd); |
1656 | return -ELOOP; | |
1657 | } | |
bc40aee0 | 1658 | if (!(nd->flags & LOOKUP_RCU)) { |
7973387a AV |
1659 | if (link->mnt == nd->path.mnt) |
1660 | mntget(link->mnt); | |
d63ff28f | 1661 | } |
626de996 AV |
1662 | error = nd_alloc_stack(nd); |
1663 | if (unlikely(error)) { | |
bc40aee0 AV |
1664 | if (error == -ECHILD) { |
1665 | if (unlikely(unlazy_link(nd, link, seq))) | |
1666 | return -ECHILD; | |
1667 | error = nd_alloc_stack(nd); | |
1668 | } | |
1669 | if (error) { | |
1670 | path_put(link); | |
1671 | return error; | |
1672 | } | |
626de996 AV |
1673 | } |
1674 | ||
ab104923 | 1675 | last = nd->stack + nd->depth++; |
1cf2665b | 1676 | last->link = *link; |
fceef393 AV |
1677 | clear_delayed_call(&last->done); |
1678 | nd->link_inode = inode; | |
0450b2d1 | 1679 | last->seq = seq; |
d63ff28f AV |
1680 | return 1; |
1681 | } | |
1682 | ||
3ddcd056 LT |
1683 | /* |
1684 | * Do we need to follow links? We _really_ want to be able | |
1685 | * to do this check without having to look at inode->i_op, | |
1686 | * so we keep a cache of "no, this doesn't need follow_link" | |
1687 | * for the common case. | |
1688 | */ | |
254cf582 | 1689 | static inline int should_follow_link(struct nameidata *nd, struct path *link, |
181548c0 AV |
1690 | int follow, |
1691 | struct inode *inode, unsigned seq) | |
3ddcd056 | 1692 | { |
d63ff28f AV |
1693 | if (likely(!d_is_symlink(link->dentry))) |
1694 | return 0; | |
1695 | if (!follow) | |
1696 | return 0; | |
a7f77542 AV |
1697 | /* make sure that d_is_symlink above matches inode */ |
1698 | if (nd->flags & LOOKUP_RCU) { | |
1699 | if (read_seqcount_retry(&link->dentry->d_seq, seq)) | |
1700 | return -ECHILD; | |
1701 | } | |
181548c0 | 1702 | return pick_link(nd, link, inode, seq); |
3ddcd056 LT |
1703 | } |
1704 | ||
4693a547 AV |
1705 | enum {WALK_GET = 1, WALK_PUT = 2}; |
1706 | ||
1707 | static int walk_component(struct nameidata *nd, int flags) | |
ce57dfc1 | 1708 | { |
caa85634 | 1709 | struct path path; |
ce57dfc1 | 1710 | struct inode *inode; |
254cf582 | 1711 | unsigned seq; |
ce57dfc1 AV |
1712 | int err; |
1713 | /* | |
1714 | * "." and ".." are special - ".." especially so because it has | |
1715 | * to be able to know about the current root directory and | |
1716 | * parent relationships. | |
1717 | */ | |
4693a547 AV |
1718 | if (unlikely(nd->last_type != LAST_NORM)) { |
1719 | err = handle_dots(nd, nd->last_type); | |
1720 | if (flags & WALK_PUT) | |
1721 | put_link(nd); | |
1722 | return err; | |
1723 | } | |
254cf582 | 1724 | err = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 1725 | if (unlikely(err <= 0)) { |
697f514d | 1726 | if (err < 0) |
f0a9ba70 | 1727 | return err; |
697f514d | 1728 | |
caa85634 | 1729 | err = lookup_slow(nd, &path); |
697f514d | 1730 | if (err < 0) |
f0a9ba70 | 1731 | return err; |
697f514d | 1732 | |
254cf582 | 1733 | seq = 0; /* we are already out of RCU mode */ |
766c4cbf | 1734 | err = -ENOENT; |
caa85634 | 1735 | if (d_is_negative(path.dentry)) |
766c4cbf | 1736 | goto out_path_put; |
d4565649 | 1737 | inode = d_backing_inode(path.dentry); |
ce57dfc1 | 1738 | } |
697f514d | 1739 | |
4693a547 AV |
1740 | if (flags & WALK_PUT) |
1741 | put_link(nd); | |
181548c0 | 1742 | err = should_follow_link(nd, &path, flags & WALK_GET, inode, seq); |
d63ff28f AV |
1743 | if (unlikely(err)) |
1744 | return err; | |
caa85634 | 1745 | path_to_nameidata(&path, nd); |
ce57dfc1 | 1746 | nd->inode = inode; |
254cf582 | 1747 | nd->seq = seq; |
ce57dfc1 | 1748 | return 0; |
697f514d MS |
1749 | |
1750 | out_path_put: | |
caa85634 | 1751 | path_to_nameidata(&path, nd); |
697f514d | 1752 | return err; |
ce57dfc1 AV |
1753 | } |
1754 | ||
bfcfaa77 LT |
1755 | /* |
1756 | * We can do the critical dentry name comparison and hashing | |
1757 | * operations one word at a time, but we are limited to: | |
1758 | * | |
1759 | * - Architectures with fast unaligned word accesses. We could | |
1760 | * do a "get_unaligned()" if this helps and is sufficiently | |
1761 | * fast. | |
1762 | * | |
bfcfaa77 LT |
1763 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1764 | * do not trap on the (extremely unlikely) case of a page | |
1765 | * crossing operation. | |
1766 | * | |
1767 | * - Furthermore, we need an efficient 64-bit compile for the | |
1768 | * 64-bit case in order to generate the "number of bytes in | |
1769 | * the final mask". Again, that could be replaced with a | |
1770 | * efficient population count instruction or similar. | |
1771 | */ | |
1772 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1773 | ||
f68e556e | 1774 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1775 | |
f68e556e | 1776 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1777 | |
1778 | static inline unsigned int fold_hash(unsigned long hash) | |
1779 | { | |
99d263d4 | 1780 | return hash_64(hash, 32); |
bfcfaa77 LT |
1781 | } |
1782 | ||
1783 | #else /* 32-bit case */ | |
1784 | ||
bfcfaa77 LT |
1785 | #define fold_hash(x) (x) |
1786 | ||
1787 | #endif | |
1788 | ||
1789 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1790 | { | |
1791 | unsigned long a, mask; | |
1792 | unsigned long hash = 0; | |
1793 | ||
1794 | for (;;) { | |
e419b4cc | 1795 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1796 | if (len < sizeof(unsigned long)) |
1797 | break; | |
1798 | hash += a; | |
f132c5be | 1799 | hash *= 9; |
bfcfaa77 LT |
1800 | name += sizeof(unsigned long); |
1801 | len -= sizeof(unsigned long); | |
1802 | if (!len) | |
1803 | goto done; | |
1804 | } | |
a5c21dce | 1805 | mask = bytemask_from_count(len); |
bfcfaa77 LT |
1806 | hash += mask & a; |
1807 | done: | |
1808 | return fold_hash(hash); | |
1809 | } | |
1810 | EXPORT_SYMBOL(full_name_hash); | |
1811 | ||
bfcfaa77 LT |
1812 | /* |
1813 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1814 | * return the "hash_len" as the result. |
bfcfaa77 | 1815 | */ |
d6bb3e90 | 1816 | static inline u64 hash_name(const char *name) |
bfcfaa77 | 1817 | { |
36126f8f LT |
1818 | unsigned long a, b, adata, bdata, mask, hash, len; |
1819 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1820 | |
1821 | hash = a = 0; | |
1822 | len = -sizeof(unsigned long); | |
1823 | do { | |
1824 | hash = (hash + a) * 9; | |
1825 | len += sizeof(unsigned long); | |
e419b4cc | 1826 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1827 | b = a ^ REPEAT_BYTE('/'); |
1828 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1829 | ||
1830 | adata = prep_zero_mask(a, adata, &constants); | |
1831 | bdata = prep_zero_mask(b, bdata, &constants); | |
1832 | ||
1833 | mask = create_zero_mask(adata | bdata); | |
1834 | ||
1835 | hash += a & zero_bytemask(mask); | |
9226b5b4 | 1836 | len += find_zero(mask); |
d6bb3e90 | 1837 | return hashlen_create(fold_hash(hash), len); |
bfcfaa77 LT |
1838 | } |
1839 | ||
1840 | #else | |
1841 | ||
0145acc2 LT |
1842 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1843 | { | |
1844 | unsigned long hash = init_name_hash(); | |
1845 | while (len--) | |
1846 | hash = partial_name_hash(*name++, hash); | |
1847 | return end_name_hash(hash); | |
1848 | } | |
ae942ae7 | 1849 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1850 | |
200e9ef7 LT |
1851 | /* |
1852 | * We know there's a real path component here of at least | |
1853 | * one character. | |
1854 | */ | |
d6bb3e90 | 1855 | static inline u64 hash_name(const char *name) |
200e9ef7 LT |
1856 | { |
1857 | unsigned long hash = init_name_hash(); | |
1858 | unsigned long len = 0, c; | |
1859 | ||
1860 | c = (unsigned char)*name; | |
1861 | do { | |
1862 | len++; | |
1863 | hash = partial_name_hash(c, hash); | |
1864 | c = (unsigned char)name[len]; | |
1865 | } while (c && c != '/'); | |
d6bb3e90 | 1866 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
1867 | } |
1868 | ||
bfcfaa77 LT |
1869 | #endif |
1870 | ||
1da177e4 LT |
1871 | /* |
1872 | * Name resolution. | |
ea3834d9 PM |
1873 | * This is the basic name resolution function, turning a pathname into |
1874 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1875 | * |
ea3834d9 PM |
1876 | * Returns 0 and nd will have valid dentry and mnt on success. |
1877 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1878 | */ |
6de88d72 | 1879 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 1880 | { |
1da177e4 | 1881 | int err; |
32cd7468 | 1882 | |
1da177e4 LT |
1883 | while (*name=='/') |
1884 | name++; | |
1885 | if (!*name) | |
9e18f10a | 1886 | return 0; |
1da177e4 | 1887 | |
1da177e4 LT |
1888 | /* At this point we know we have a real path component. */ |
1889 | for(;;) { | |
d6bb3e90 | 1890 | u64 hash_len; |
fe479a58 | 1891 | int type; |
1da177e4 | 1892 | |
52094c8a | 1893 | err = may_lookup(nd); |
1da177e4 | 1894 | if (err) |
3595e234 | 1895 | return err; |
1da177e4 | 1896 | |
d6bb3e90 | 1897 | hash_len = hash_name(name); |
1da177e4 | 1898 | |
fe479a58 | 1899 | type = LAST_NORM; |
d6bb3e90 | 1900 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 1901 | case 2: |
200e9ef7 | 1902 | if (name[1] == '.') { |
fe479a58 | 1903 | type = LAST_DOTDOT; |
16c2cd71 AV |
1904 | nd->flags |= LOOKUP_JUMPED; |
1905 | } | |
fe479a58 AV |
1906 | break; |
1907 | case 1: | |
1908 | type = LAST_DOT; | |
1909 | } | |
5a202bcd AV |
1910 | if (likely(type == LAST_NORM)) { |
1911 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1912 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 1913 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 1914 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 1915 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 1916 | if (err < 0) |
3595e234 | 1917 | return err; |
d6bb3e90 LT |
1918 | hash_len = this.hash_len; |
1919 | name = this.name; | |
5a202bcd AV |
1920 | } |
1921 | } | |
fe479a58 | 1922 | |
d6bb3e90 LT |
1923 | nd->last.hash_len = hash_len; |
1924 | nd->last.name = name; | |
5f4a6a69 AV |
1925 | nd->last_type = type; |
1926 | ||
d6bb3e90 LT |
1927 | name += hashlen_len(hash_len); |
1928 | if (!*name) | |
bdf6cbf1 | 1929 | goto OK; |
200e9ef7 LT |
1930 | /* |
1931 | * If it wasn't NUL, we know it was '/'. Skip that | |
1932 | * slash, and continue until no more slashes. | |
1933 | */ | |
1934 | do { | |
d6bb3e90 LT |
1935 | name++; |
1936 | } while (unlikely(*name == '/')); | |
8620c238 AV |
1937 | if (unlikely(!*name)) { |
1938 | OK: | |
368ee9ba | 1939 | /* pathname body, done */ |
8620c238 AV |
1940 | if (!nd->depth) |
1941 | return 0; | |
1942 | name = nd->stack[nd->depth - 1].name; | |
368ee9ba | 1943 | /* trailing symlink, done */ |
8620c238 AV |
1944 | if (!name) |
1945 | return 0; | |
1946 | /* last component of nested symlink */ | |
4693a547 | 1947 | err = walk_component(nd, WALK_GET | WALK_PUT); |
8620c238 | 1948 | } else { |
4693a547 | 1949 | err = walk_component(nd, WALK_GET); |
8620c238 | 1950 | } |
ce57dfc1 | 1951 | if (err < 0) |
3595e234 | 1952 | return err; |
1da177e4 | 1953 | |
ce57dfc1 | 1954 | if (err) { |
626de996 | 1955 | const char *s = get_link(nd); |
5a460275 | 1956 | |
a1c83681 | 1957 | if (IS_ERR(s)) |
3595e234 | 1958 | return PTR_ERR(s); |
d40bcc09 AV |
1959 | err = 0; |
1960 | if (unlikely(!s)) { | |
1961 | /* jumped */ | |
b9ff4429 | 1962 | put_link(nd); |
d40bcc09 | 1963 | } else { |
fab51e8a AV |
1964 | nd->stack[nd->depth - 1].name = name; |
1965 | name = s; | |
1966 | continue; | |
d40bcc09 | 1967 | } |
31e6b01f | 1968 | } |
97242f99 AV |
1969 | if (unlikely(!d_can_lookup(nd->path.dentry))) { |
1970 | if (nd->flags & LOOKUP_RCU) { | |
1971 | if (unlazy_walk(nd, NULL, 0)) | |
1972 | return -ECHILD; | |
1973 | } | |
3595e234 | 1974 | return -ENOTDIR; |
97242f99 | 1975 | } |
1da177e4 | 1976 | } |
1da177e4 LT |
1977 | } |
1978 | ||
c8a53ee5 | 1979 | static const char *path_init(struct nameidata *nd, unsigned flags) |
31e6b01f NP |
1980 | { |
1981 | int retval = 0; | |
c8a53ee5 | 1982 | const char *s = nd->name->name; |
31e6b01f NP |
1983 | |
1984 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
980f3ea2 | 1985 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 1986 | nd->depth = 0; |
5b6ca027 | 1987 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
1988 | struct dentry *root = nd->root.dentry; |
1989 | struct inode *inode = root->d_inode; | |
fd2f7cb5 | 1990 | if (*s) { |
44b1d530 | 1991 | if (!d_can_lookup(root)) |
368ee9ba | 1992 | return ERR_PTR(-ENOTDIR); |
73d049a4 AV |
1993 | retval = inode_permission(inode, MAY_EXEC); |
1994 | if (retval) | |
368ee9ba | 1995 | return ERR_PTR(retval); |
73d049a4 | 1996 | } |
5b6ca027 AV |
1997 | nd->path = nd->root; |
1998 | nd->inode = inode; | |
1999 | if (flags & LOOKUP_RCU) { | |
8b61e74f | 2000 | rcu_read_lock(); |
5b6ca027 | 2001 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
8f47a016 | 2002 | nd->root_seq = nd->seq; |
48a066e7 | 2003 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
2004 | } else { |
2005 | path_get(&nd->path); | |
2006 | } | |
368ee9ba | 2007 | return s; |
5b6ca027 AV |
2008 | } |
2009 | ||
31e6b01f | 2010 | nd->root.mnt = NULL; |
248fb5b9 AV |
2011 | nd->path.mnt = NULL; |
2012 | nd->path.dentry = NULL; | |
31e6b01f | 2013 | |
48a066e7 | 2014 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 2015 | if (*s == '/') { |
9e6697e2 | 2016 | if (flags & LOOKUP_RCU) |
8b61e74f | 2017 | rcu_read_lock(); |
9e6697e2 | 2018 | set_root(nd); |
248fb5b9 | 2019 | if (likely(!nd_jump_root(nd))) |
ef55d917 | 2020 | return s; |
248fb5b9 | 2021 | nd->root.mnt = NULL; |
ef55d917 AV |
2022 | rcu_read_unlock(); |
2023 | return ERR_PTR(-ECHILD); | |
c8a53ee5 | 2024 | } else if (nd->dfd == AT_FDCWD) { |
e41f7d4e AV |
2025 | if (flags & LOOKUP_RCU) { |
2026 | struct fs_struct *fs = current->fs; | |
2027 | unsigned seq; | |
31e6b01f | 2028 | |
8b61e74f | 2029 | rcu_read_lock(); |
c28cc364 | 2030 | |
e41f7d4e AV |
2031 | do { |
2032 | seq = read_seqcount_begin(&fs->seq); | |
2033 | nd->path = fs->pwd; | |
ef55d917 | 2034 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e AV |
2035 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
2036 | } while (read_seqcount_retry(&fs->seq, seq)); | |
2037 | } else { | |
2038 | get_fs_pwd(current->fs, &nd->path); | |
ef55d917 | 2039 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2040 | } |
ef55d917 | 2041 | return s; |
31e6b01f | 2042 | } else { |
582aa64a | 2043 | /* Caller must check execute permissions on the starting path component */ |
c8a53ee5 | 2044 | struct fd f = fdget_raw(nd->dfd); |
31e6b01f NP |
2045 | struct dentry *dentry; |
2046 | ||
2903ff01 | 2047 | if (!f.file) |
368ee9ba | 2048 | return ERR_PTR(-EBADF); |
31e6b01f | 2049 | |
2903ff01 | 2050 | dentry = f.file->f_path.dentry; |
31e6b01f | 2051 | |
fd2f7cb5 | 2052 | if (*s) { |
44b1d530 | 2053 | if (!d_can_lookup(dentry)) { |
2903ff01 | 2054 | fdput(f); |
368ee9ba | 2055 | return ERR_PTR(-ENOTDIR); |
2903ff01 | 2056 | } |
f52e0c11 | 2057 | } |
31e6b01f | 2058 | |
2903ff01 | 2059 | nd->path = f.file->f_path; |
e41f7d4e | 2060 | if (flags & LOOKUP_RCU) { |
8b61e74f | 2061 | rcu_read_lock(); |
34a26b99 AV |
2062 | nd->inode = nd->path.dentry->d_inode; |
2063 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 2064 | } else { |
2903ff01 | 2065 | path_get(&nd->path); |
34a26b99 | 2066 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2067 | } |
34a26b99 | 2068 | fdput(f); |
368ee9ba | 2069 | return s; |
31e6b01f | 2070 | } |
9b4a9b14 AV |
2071 | } |
2072 | ||
3bdba28b | 2073 | static const char *trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
2074 | { |
2075 | const char *s; | |
fec2fa24 | 2076 | int error = may_follow_link(nd); |
deb106c6 | 2077 | if (unlikely(error)) |
3bdba28b | 2078 | return ERR_PTR(error); |
95fa25d9 | 2079 | nd->flags |= LOOKUP_PARENT; |
fab51e8a | 2080 | nd->stack[0].name = NULL; |
3b2e7f75 | 2081 | s = get_link(nd); |
deb106c6 | 2082 | return s ? s : ""; |
95fa25d9 AV |
2083 | } |
2084 | ||
caa85634 | 2085 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2086 | { |
2087 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2088 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2089 | ||
2090 | nd->flags &= ~LOOKUP_PARENT; | |
deb106c6 | 2091 | return walk_component(nd, |
4693a547 AV |
2092 | nd->flags & LOOKUP_FOLLOW |
2093 | ? nd->depth | |
2094 | ? WALK_PUT | WALK_GET | |
2095 | : WALK_GET | |
2096 | : 0); | |
bd92d7fe AV |
2097 | } |
2098 | ||
9b4a9b14 | 2099 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2100 | static int path_lookupat(struct nameidata *nd, unsigned flags, struct path *path) |
9b4a9b14 | 2101 | { |
c8a53ee5 | 2102 | const char *s = path_init(nd, flags); |
bd92d7fe | 2103 | int err; |
31e6b01f | 2104 | |
368ee9ba AV |
2105 | if (IS_ERR(s)) |
2106 | return PTR_ERR(s); | |
3bdba28b AV |
2107 | while (!(err = link_path_walk(s, nd)) |
2108 | && ((err = lookup_last(nd)) > 0)) { | |
2109 | s = trailing_symlink(nd); | |
2110 | if (IS_ERR(s)) { | |
2111 | err = PTR_ERR(s); | |
2112 | break; | |
bd92d7fe AV |
2113 | } |
2114 | } | |
9f1fafee AV |
2115 | if (!err) |
2116 | err = complete_walk(nd); | |
bd92d7fe | 2117 | |
deb106c6 AV |
2118 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2119 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2120 | err = -ENOTDIR; |
625b6d10 AV |
2121 | if (!err) { |
2122 | *path = nd->path; | |
2123 | nd->path.mnt = NULL; | |
2124 | nd->path.dentry = NULL; | |
2125 | } | |
2126 | terminate_walk(nd); | |
bd92d7fe | 2127 | return err; |
ee0827cd | 2128 | } |
31e6b01f | 2129 | |
625b6d10 | 2130 | static int filename_lookup(int dfd, struct filename *name, unsigned flags, |
9ad1aaa6 | 2131 | struct path *path, struct path *root) |
ee0827cd | 2132 | { |
894bc8c4 | 2133 | int retval; |
9883d185 | 2134 | struct nameidata nd; |
abc9f5be AV |
2135 | if (IS_ERR(name)) |
2136 | return PTR_ERR(name); | |
9ad1aaa6 AV |
2137 | if (unlikely(root)) { |
2138 | nd.root = *root; | |
2139 | flags |= LOOKUP_ROOT; | |
2140 | } | |
9883d185 | 2141 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2142 | retval = path_lookupat(&nd, flags | LOOKUP_RCU, path); |
ee0827cd | 2143 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2144 | retval = path_lookupat(&nd, flags, path); |
ee0827cd | 2145 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2146 | retval = path_lookupat(&nd, flags | LOOKUP_REVAL, path); |
31e6b01f | 2147 | |
f78570dd | 2148 | if (likely(!retval)) |
625b6d10 | 2149 | audit_inode(name, path->dentry, flags & LOOKUP_PARENT); |
9883d185 | 2150 | restore_nameidata(); |
e4bd1c1a | 2151 | putname(name); |
170aa3d0 | 2152 | return retval; |
1da177e4 LT |
2153 | } |
2154 | ||
8bcb77fa | 2155 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2156 | static int path_parentat(struct nameidata *nd, unsigned flags, |
391172c4 | 2157 | struct path *parent) |
8bcb77fa | 2158 | { |
c8a53ee5 | 2159 | const char *s = path_init(nd, flags); |
368ee9ba AV |
2160 | int err; |
2161 | if (IS_ERR(s)) | |
2162 | return PTR_ERR(s); | |
2163 | err = link_path_walk(s, nd); | |
8bcb77fa AV |
2164 | if (!err) |
2165 | err = complete_walk(nd); | |
391172c4 AV |
2166 | if (!err) { |
2167 | *parent = nd->path; | |
2168 | nd->path.mnt = NULL; | |
2169 | nd->path.dentry = NULL; | |
2170 | } | |
2171 | terminate_walk(nd); | |
8bcb77fa AV |
2172 | return err; |
2173 | } | |
2174 | ||
5c31b6ce | 2175 | static struct filename *filename_parentat(int dfd, struct filename *name, |
391172c4 AV |
2176 | unsigned int flags, struct path *parent, |
2177 | struct qstr *last, int *type) | |
8bcb77fa AV |
2178 | { |
2179 | int retval; | |
9883d185 | 2180 | struct nameidata nd; |
8bcb77fa | 2181 | |
5c31b6ce AV |
2182 | if (IS_ERR(name)) |
2183 | return name; | |
9883d185 | 2184 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2185 | retval = path_parentat(&nd, flags | LOOKUP_RCU, parent); |
8bcb77fa | 2186 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2187 | retval = path_parentat(&nd, flags, parent); |
8bcb77fa | 2188 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2189 | retval = path_parentat(&nd, flags | LOOKUP_REVAL, parent); |
391172c4 AV |
2190 | if (likely(!retval)) { |
2191 | *last = nd.last; | |
2192 | *type = nd.last_type; | |
2193 | audit_inode(name, parent->dentry, LOOKUP_PARENT); | |
5c31b6ce AV |
2194 | } else { |
2195 | putname(name); | |
2196 | name = ERR_PTR(retval); | |
391172c4 | 2197 | } |
9883d185 | 2198 | restore_nameidata(); |
5c31b6ce | 2199 | return name; |
8bcb77fa AV |
2200 | } |
2201 | ||
79714f72 AV |
2202 | /* does lookup, returns the object with parent locked */ |
2203 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2204 | { |
5c31b6ce AV |
2205 | struct filename *filename; |
2206 | struct dentry *d; | |
391172c4 AV |
2207 | struct qstr last; |
2208 | int type; | |
51689104 | 2209 | |
5c31b6ce AV |
2210 | filename = filename_parentat(AT_FDCWD, getname_kernel(name), 0, path, |
2211 | &last, &type); | |
51689104 PM |
2212 | if (IS_ERR(filename)) |
2213 | return ERR_CAST(filename); | |
5c31b6ce | 2214 | if (unlikely(type != LAST_NORM)) { |
391172c4 | 2215 | path_put(path); |
5c31b6ce AV |
2216 | putname(filename); |
2217 | return ERR_PTR(-EINVAL); | |
79714f72 | 2218 | } |
5955102c | 2219 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 2220 | d = __lookup_hash(&last, path->dentry, 0); |
79714f72 | 2221 | if (IS_ERR(d)) { |
5955102c | 2222 | inode_unlock(path->dentry->d_inode); |
391172c4 | 2223 | path_put(path); |
79714f72 | 2224 | } |
51689104 | 2225 | putname(filename); |
79714f72 | 2226 | return d; |
5590ff0d UD |
2227 | } |
2228 | ||
d1811465 AV |
2229 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2230 | { | |
abc9f5be AV |
2231 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2232 | flags, path, NULL); | |
d1811465 | 2233 | } |
4d359507 | 2234 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2235 | |
16f18200 JJS |
2236 | /** |
2237 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2238 | * @dentry: pointer to dentry of the base directory | |
2239 | * @mnt: pointer to vfs mount of the base directory | |
2240 | * @name: pointer to file name | |
2241 | * @flags: lookup flags | |
e0a01249 | 2242 | * @path: pointer to struct path to fill |
16f18200 JJS |
2243 | */ |
2244 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2245 | const char *name, unsigned int flags, | |
e0a01249 | 2246 | struct path *path) |
16f18200 | 2247 | { |
9ad1aaa6 | 2248 | struct path root = {.mnt = mnt, .dentry = dentry}; |
9ad1aaa6 | 2249 | /* the first argument of filename_lookup() is ignored with root */ |
abc9f5be AV |
2250 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2251 | flags , path, &root); | |
16f18200 | 2252 | } |
4d359507 | 2253 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2254 | |
eead1911 | 2255 | /** |
a6b91919 | 2256 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2257 | * @name: pathname component to lookup |
2258 | * @base: base directory to lookup from | |
2259 | * @len: maximum length @len should be interpreted to | |
2260 | * | |
a6b91919 | 2261 | * Note that this routine is purely a helper for filesystem usage and should |
9e7543e9 | 2262 | * not be called by generic code. |
bbddca8e N |
2263 | * |
2264 | * The caller must hold base->i_mutex. | |
eead1911 | 2265 | */ |
057f6c01 JM |
2266 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2267 | { | |
057f6c01 | 2268 | struct qstr this; |
6a96ba54 | 2269 | unsigned int c; |
cda309de | 2270 | int err; |
057f6c01 | 2271 | |
5955102c | 2272 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); |
2f9092e1 | 2273 | |
6a96ba54 AV |
2274 | this.name = name; |
2275 | this.len = len; | |
0145acc2 | 2276 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2277 | if (!len) |
2278 | return ERR_PTR(-EACCES); | |
2279 | ||
21d8a15a AV |
2280 | if (unlikely(name[0] == '.')) { |
2281 | if (len < 2 || (len == 2 && name[1] == '.')) | |
2282 | return ERR_PTR(-EACCES); | |
2283 | } | |
2284 | ||
6a96ba54 AV |
2285 | while (len--) { |
2286 | c = *(const unsigned char *)name++; | |
2287 | if (c == '/' || c == '\0') | |
2288 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2289 | } |
5a202bcd AV |
2290 | /* |
2291 | * See if the low-level filesystem might want | |
2292 | * to use its own hash.. | |
2293 | */ | |
2294 | if (base->d_flags & DCACHE_OP_HASH) { | |
da53be12 | 2295 | int err = base->d_op->d_hash(base, &this); |
5a202bcd AV |
2296 | if (err < 0) |
2297 | return ERR_PTR(err); | |
2298 | } | |
eead1911 | 2299 | |
cda309de MS |
2300 | err = inode_permission(base->d_inode, MAY_EXEC); |
2301 | if (err) | |
2302 | return ERR_PTR(err); | |
2303 | ||
72bd866a | 2304 | return __lookup_hash(&this, base, 0); |
057f6c01 | 2305 | } |
4d359507 | 2306 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2307 | |
bbddca8e N |
2308 | /** |
2309 | * lookup_one_len_unlocked - filesystem helper to lookup single pathname component | |
2310 | * @name: pathname component to lookup | |
2311 | * @base: base directory to lookup from | |
2312 | * @len: maximum length @len should be interpreted to | |
2313 | * | |
2314 | * Note that this routine is purely a helper for filesystem usage and should | |
2315 | * not be called by generic code. | |
2316 | * | |
2317 | * Unlike lookup_one_len, it should be called without the parent | |
2318 | * i_mutex held, and will take the i_mutex itself if necessary. | |
2319 | */ | |
2320 | struct dentry *lookup_one_len_unlocked(const char *name, | |
2321 | struct dentry *base, int len) | |
2322 | { | |
2323 | struct qstr this; | |
2324 | unsigned int c; | |
2325 | int err; | |
2326 | struct dentry *ret; | |
2327 | ||
2328 | this.name = name; | |
2329 | this.len = len; | |
2330 | this.hash = full_name_hash(name, len); | |
2331 | if (!len) | |
2332 | return ERR_PTR(-EACCES); | |
2333 | ||
2334 | if (unlikely(name[0] == '.')) { | |
2335 | if (len < 2 || (len == 2 && name[1] == '.')) | |
2336 | return ERR_PTR(-EACCES); | |
2337 | } | |
2338 | ||
2339 | while (len--) { | |
2340 | c = *(const unsigned char *)name++; | |
2341 | if (c == '/' || c == '\0') | |
2342 | return ERR_PTR(-EACCES); | |
2343 | } | |
2344 | /* | |
2345 | * See if the low-level filesystem might want | |
2346 | * to use its own hash.. | |
2347 | */ | |
2348 | if (base->d_flags & DCACHE_OP_HASH) { | |
2349 | int err = base->d_op->d_hash(base, &this); | |
2350 | if (err < 0) | |
2351 | return ERR_PTR(err); | |
2352 | } | |
2353 | ||
2354 | err = inode_permission(base->d_inode, MAY_EXEC); | |
2355 | if (err) | |
2356 | return ERR_PTR(err); | |
2357 | ||
2358 | /* | |
2359 | * __d_lookup() is used to try to get a quick answer and avoid the | |
2360 | * mutex. A false-negative does no harm. | |
2361 | */ | |
2362 | ret = __d_lookup(base, &this); | |
2363 | if (ret && unlikely(ret->d_flags & DCACHE_OP_REVALIDATE)) { | |
2364 | dput(ret); | |
2365 | ret = NULL; | |
2366 | } | |
2367 | if (ret) | |
2368 | return ret; | |
2369 | ||
5955102c | 2370 | inode_lock(base->d_inode); |
bbddca8e | 2371 | ret = __lookup_hash(&this, base, 0); |
5955102c | 2372 | inode_unlock(base->d_inode); |
bbddca8e N |
2373 | return ret; |
2374 | } | |
2375 | EXPORT_SYMBOL(lookup_one_len_unlocked); | |
2376 | ||
1fa1e7f6 AW |
2377 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2378 | struct path *path, int *empty) | |
1da177e4 | 2379 | { |
abc9f5be AV |
2380 | return filename_lookup(dfd, getname_flags(name, flags, empty), |
2381 | flags, path, NULL); | |
1da177e4 | 2382 | } |
b853a161 | 2383 | EXPORT_SYMBOL(user_path_at_empty); |
1fa1e7f6 | 2384 | |
873f1eed JL |
2385 | /* |
2386 | * NB: most callers don't do anything directly with the reference to the | |
2387 | * to struct filename, but the nd->last pointer points into the name string | |
2388 | * allocated by getname. So we must hold the reference to it until all | |
2389 | * path-walking is complete. | |
2390 | */ | |
a2ec4a2d | 2391 | static inline struct filename * |
f5beed75 AV |
2392 | user_path_parent(int dfd, const char __user *path, |
2393 | struct path *parent, | |
2394 | struct qstr *last, | |
2395 | int *type, | |
9e790bd6 | 2396 | unsigned int flags) |
2ad94ae6 | 2397 | { |
9e790bd6 | 2398 | /* only LOOKUP_REVAL is allowed in extra flags */ |
5c31b6ce AV |
2399 | return filename_parentat(dfd, getname(path), flags & LOOKUP_REVAL, |
2400 | parent, last, type); | |
2ad94ae6 AV |
2401 | } |
2402 | ||
8033426e | 2403 | /** |
197df04c | 2404 | * mountpoint_last - look up last component for umount |
8033426e JL |
2405 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
2406 | * @path: pointer to container for result | |
2407 | * | |
2408 | * This is a special lookup_last function just for umount. In this case, we | |
2409 | * need to resolve the path without doing any revalidation. | |
2410 | * | |
2411 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2412 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2413 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2414 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2415 | * bogus and it doesn't exist. | |
2416 | * | |
2417 | * Returns: | |
2418 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
2419 | * lookup found a negative dentry. The nd->path reference will also be | |
2420 | * put in this case. | |
2421 | * | |
2422 | * 0: if we successfully resolved nd->path and found it to not to be a | |
2423 | * symlink that needs to be followed. "path" will also be populated. | |
2424 | * The nd->path reference will also be put. | |
2425 | * | |
2426 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
2427 | * that needs to be followed. "path" will be populated with the path | |
2428 | * to the link, and nd->path will *not* be put. | |
2429 | */ | |
2430 | static int | |
197df04c | 2431 | mountpoint_last(struct nameidata *nd, struct path *path) |
8033426e JL |
2432 | { |
2433 | int error = 0; | |
2434 | struct dentry *dentry; | |
2435 | struct dentry *dir = nd->path.dentry; | |
2436 | ||
35759521 AV |
2437 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2438 | if (nd->flags & LOOKUP_RCU) { | |
6e9918b7 | 2439 | if (unlazy_walk(nd, NULL, 0)) |
deb106c6 | 2440 | return -ECHILD; |
8033426e JL |
2441 | } |
2442 | ||
2443 | nd->flags &= ~LOOKUP_PARENT; | |
2444 | ||
2445 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2446 | error = handle_dots(nd, nd->last_type); | |
35759521 | 2447 | if (error) |
deb106c6 | 2448 | return error; |
35759521 AV |
2449 | dentry = dget(nd->path.dentry); |
2450 | goto done; | |
8033426e JL |
2451 | } |
2452 | ||
5955102c | 2453 | inode_lock(dir->d_inode); |
8033426e JL |
2454 | dentry = d_lookup(dir, &nd->last); |
2455 | if (!dentry) { | |
2456 | /* | |
2457 | * No cached dentry. Mounted dentries are pinned in the cache, | |
2458 | * so that means that this dentry is probably a symlink or the | |
2459 | * path doesn't actually point to a mounted dentry. | |
2460 | */ | |
2461 | dentry = d_alloc(dir, &nd->last); | |
2462 | if (!dentry) { | |
5955102c | 2463 | inode_unlock(dir->d_inode); |
deb106c6 | 2464 | return -ENOMEM; |
8033426e | 2465 | } |
35759521 | 2466 | dentry = lookup_real(dir->d_inode, dentry, nd->flags); |
bcceeeba | 2467 | if (IS_ERR(dentry)) { |
5955102c | 2468 | inode_unlock(dir->d_inode); |
deb106c6 | 2469 | return PTR_ERR(dentry); |
bcceeeba | 2470 | } |
8033426e | 2471 | } |
5955102c | 2472 | inode_unlock(dir->d_inode); |
8033426e | 2473 | |
35759521 | 2474 | done: |
698934df | 2475 | if (d_is_negative(dentry)) { |
35759521 | 2476 | dput(dentry); |
deb106c6 | 2477 | return -ENOENT; |
8033426e | 2478 | } |
191d7f73 AV |
2479 | if (nd->depth) |
2480 | put_link(nd); | |
35759521 | 2481 | path->dentry = dentry; |
295dc39d | 2482 | path->mnt = nd->path.mnt; |
181548c0 AV |
2483 | error = should_follow_link(nd, path, nd->flags & LOOKUP_FOLLOW, |
2484 | d_backing_inode(dentry), 0); | |
deb106c6 | 2485 | if (unlikely(error)) |
d63ff28f | 2486 | return error; |
295dc39d | 2487 | mntget(path->mnt); |
35759521 | 2488 | follow_mount(path); |
deb106c6 | 2489 | return 0; |
8033426e JL |
2490 | } |
2491 | ||
2492 | /** | |
197df04c | 2493 | * path_mountpoint - look up a path to be umounted |
2a78b857 | 2494 | * @nd: lookup context |
8033426e | 2495 | * @flags: lookup flags |
c8a53ee5 | 2496 | * @path: pointer to container for result |
8033426e JL |
2497 | * |
2498 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2499 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2500 | */ |
2501 | static int | |
c8a53ee5 | 2502 | path_mountpoint(struct nameidata *nd, unsigned flags, struct path *path) |
8033426e | 2503 | { |
c8a53ee5 | 2504 | const char *s = path_init(nd, flags); |
368ee9ba AV |
2505 | int err; |
2506 | if (IS_ERR(s)) | |
2507 | return PTR_ERR(s); | |
3bdba28b AV |
2508 | while (!(err = link_path_walk(s, nd)) && |
2509 | (err = mountpoint_last(nd, path)) > 0) { | |
2510 | s = trailing_symlink(nd); | |
2511 | if (IS_ERR(s)) { | |
2512 | err = PTR_ERR(s); | |
8033426e | 2513 | break; |
3bdba28b | 2514 | } |
8033426e | 2515 | } |
deb106c6 | 2516 | terminate_walk(nd); |
8033426e JL |
2517 | return err; |
2518 | } | |
2519 | ||
2d864651 | 2520 | static int |
668696dc | 2521 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2522 | unsigned int flags) |
2523 | { | |
9883d185 | 2524 | struct nameidata nd; |
cbaab2db | 2525 | int error; |
668696dc AV |
2526 | if (IS_ERR(name)) |
2527 | return PTR_ERR(name); | |
9883d185 | 2528 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2529 | error = path_mountpoint(&nd, flags | LOOKUP_RCU, path); |
2d864651 | 2530 | if (unlikely(error == -ECHILD)) |
c8a53ee5 | 2531 | error = path_mountpoint(&nd, flags, path); |
2d864651 | 2532 | if (unlikely(error == -ESTALE)) |
c8a53ee5 | 2533 | error = path_mountpoint(&nd, flags | LOOKUP_REVAL, path); |
2d864651 | 2534 | if (likely(!error)) |
668696dc | 2535 | audit_inode(name, path->dentry, 0); |
9883d185 | 2536 | restore_nameidata(); |
668696dc | 2537 | putname(name); |
2d864651 AV |
2538 | return error; |
2539 | } | |
2540 | ||
8033426e | 2541 | /** |
197df04c | 2542 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2543 | * @dfd: directory file descriptor |
2544 | * @name: pathname from userland | |
2545 | * @flags: lookup flags | |
2546 | * @path: pointer to container to hold result | |
2547 | * | |
2548 | * A umount is a special case for path walking. We're not actually interested | |
2549 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2550 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2551 | * and avoid revalidating the last component. | |
2552 | * | |
2553 | * Returns 0 and populates "path" on success. | |
2554 | */ | |
2555 | int | |
197df04c | 2556 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2557 | struct path *path) |
2558 | { | |
cbaab2db | 2559 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2560 | } |
2561 | ||
2d864651 AV |
2562 | int |
2563 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2564 | unsigned int flags) | |
2565 | { | |
cbaab2db | 2566 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2567 | } |
2568 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2569 | ||
cbdf35bc | 2570 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2571 | { |
8e96e3b7 | 2572 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2573 | |
8e96e3b7 | 2574 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2575 | return 0; |
8e96e3b7 | 2576 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2577 | return 0; |
23adbe12 | 2578 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2579 | } |
cbdf35bc | 2580 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2581 | |
2582 | /* | |
2583 | * Check whether we can remove a link victim from directory dir, check | |
2584 | * whether the type of victim is right. | |
2585 | * 1. We can't do it if dir is read-only (done in permission()) | |
2586 | * 2. We should have write and exec permissions on dir | |
2587 | * 3. We can't remove anything from append-only dir | |
2588 | * 4. We can't do anything with immutable dir (done in permission()) | |
2589 | * 5. If the sticky bit on dir is set we should either | |
2590 | * a. be owner of dir, or | |
2591 | * b. be owner of victim, or | |
2592 | * c. have CAP_FOWNER capability | |
2593 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2594 | * links pointing to it. | |
2595 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2596 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2597 | * 9. We can't remove a root or mountpoint. | |
2598 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2599 | * nfs_async_unlink(). | |
2600 | */ | |
b18825a7 | 2601 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2602 | { |
63afdfc7 | 2603 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2604 | int error; |
2605 | ||
b18825a7 | 2606 | if (d_is_negative(victim)) |
1da177e4 | 2607 | return -ENOENT; |
b18825a7 | 2608 | BUG_ON(!inode); |
1da177e4 LT |
2609 | |
2610 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 2611 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2612 | |
f419a2e3 | 2613 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2614 | if (error) |
2615 | return error; | |
2616 | if (IS_APPEND(dir)) | |
2617 | return -EPERM; | |
b18825a7 DH |
2618 | |
2619 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
2620 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode)) | |
1da177e4 LT |
2621 | return -EPERM; |
2622 | if (isdir) { | |
44b1d530 | 2623 | if (!d_is_dir(victim)) |
1da177e4 LT |
2624 | return -ENOTDIR; |
2625 | if (IS_ROOT(victim)) | |
2626 | return -EBUSY; | |
44b1d530 | 2627 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2628 | return -EISDIR; |
2629 | if (IS_DEADDIR(dir)) | |
2630 | return -ENOENT; | |
2631 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2632 | return -EBUSY; | |
2633 | return 0; | |
2634 | } | |
2635 | ||
2636 | /* Check whether we can create an object with dentry child in directory | |
2637 | * dir. | |
2638 | * 1. We can't do it if child already exists (open has special treatment for | |
2639 | * this case, but since we are inlined it's OK) | |
2640 | * 2. We can't do it if dir is read-only (done in permission()) | |
2641 | * 3. We should have write and exec permissions on dir | |
2642 | * 4. We can't do it if dir is immutable (done in permission()) | |
2643 | */ | |
a95164d9 | 2644 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2645 | { |
14e972b4 | 2646 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2647 | if (child->d_inode) |
2648 | return -EEXIST; | |
2649 | if (IS_DEADDIR(dir)) | |
2650 | return -ENOENT; | |
f419a2e3 | 2651 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2652 | } |
2653 | ||
1da177e4 LT |
2654 | /* |
2655 | * p1 and p2 should be directories on the same fs. | |
2656 | */ | |
2657 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2658 | { | |
2659 | struct dentry *p; | |
2660 | ||
2661 | if (p1 == p2) { | |
5955102c | 2662 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
1da177e4 LT |
2663 | return NULL; |
2664 | } | |
2665 | ||
a11f3a05 | 2666 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2667 | |
e2761a11 OH |
2668 | p = d_ancestor(p2, p1); |
2669 | if (p) { | |
5955102c AV |
2670 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT); |
2671 | inode_lock_nested(p1->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2672 | return p; |
1da177e4 LT |
2673 | } |
2674 | ||
e2761a11 OH |
2675 | p = d_ancestor(p1, p2); |
2676 | if (p) { | |
5955102c AV |
2677 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2678 | inode_lock_nested(p2->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2679 | return p; |
1da177e4 LT |
2680 | } |
2681 | ||
5955102c AV |
2682 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2683 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT2); | |
1da177e4 LT |
2684 | return NULL; |
2685 | } | |
4d359507 | 2686 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2687 | |
2688 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2689 | { | |
5955102c | 2690 | inode_unlock(p1->d_inode); |
1da177e4 | 2691 | if (p1 != p2) { |
5955102c | 2692 | inode_unlock(p2->d_inode); |
a11f3a05 | 2693 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2694 | } |
2695 | } | |
4d359507 | 2696 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2697 | |
4acdaf27 | 2698 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2699 | bool want_excl) |
1da177e4 | 2700 | { |
a95164d9 | 2701 | int error = may_create(dir, dentry); |
1da177e4 LT |
2702 | if (error) |
2703 | return error; | |
2704 | ||
acfa4380 | 2705 | if (!dir->i_op->create) |
1da177e4 LT |
2706 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2707 | mode &= S_IALLUGO; | |
2708 | mode |= S_IFREG; | |
2709 | error = security_inode_create(dir, dentry, mode); | |
2710 | if (error) | |
2711 | return error; | |
312b63fb | 2712 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2713 | if (!error) |
f38aa942 | 2714 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2715 | return error; |
2716 | } | |
4d359507 | 2717 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2718 | |
73d049a4 | 2719 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2720 | { |
3fb64190 | 2721 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2722 | struct inode *inode = dentry->d_inode; |
2723 | int error; | |
2724 | ||
2725 | if (!inode) | |
2726 | return -ENOENT; | |
2727 | ||
c8fe8f30 CH |
2728 | switch (inode->i_mode & S_IFMT) { |
2729 | case S_IFLNK: | |
1da177e4 | 2730 | return -ELOOP; |
c8fe8f30 CH |
2731 | case S_IFDIR: |
2732 | if (acc_mode & MAY_WRITE) | |
2733 | return -EISDIR; | |
2734 | break; | |
2735 | case S_IFBLK: | |
2736 | case S_IFCHR: | |
3fb64190 | 2737 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2738 | return -EACCES; |
c8fe8f30 CH |
2739 | /*FALLTHRU*/ |
2740 | case S_IFIFO: | |
2741 | case S_IFSOCK: | |
1da177e4 | 2742 | flag &= ~O_TRUNC; |
c8fe8f30 | 2743 | break; |
4a3fd211 | 2744 | } |
b41572e9 | 2745 | |
62fb4a15 | 2746 | error = inode_permission(inode, MAY_OPEN | acc_mode); |
b41572e9 DH |
2747 | if (error) |
2748 | return error; | |
6146f0d5 | 2749 | |
1da177e4 LT |
2750 | /* |
2751 | * An append-only file must be opened in append mode for writing. | |
2752 | */ | |
2753 | if (IS_APPEND(inode)) { | |
8737c930 | 2754 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2755 | return -EPERM; |
1da177e4 | 2756 | if (flag & O_TRUNC) |
7715b521 | 2757 | return -EPERM; |
1da177e4 LT |
2758 | } |
2759 | ||
2760 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2761 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2762 | return -EPERM; |
1da177e4 | 2763 | |
f3c7691e | 2764 | return 0; |
7715b521 | 2765 | } |
1da177e4 | 2766 | |
e1181ee6 | 2767 | static int handle_truncate(struct file *filp) |
7715b521 | 2768 | { |
e1181ee6 | 2769 | struct path *path = &filp->f_path; |
7715b521 AV |
2770 | struct inode *inode = path->dentry->d_inode; |
2771 | int error = get_write_access(inode); | |
2772 | if (error) | |
2773 | return error; | |
2774 | /* | |
2775 | * Refuse to truncate files with mandatory locks held on them. | |
2776 | */ | |
d7a06983 | 2777 | error = locks_verify_locked(filp); |
7715b521 | 2778 | if (!error) |
ea0d3ab2 | 2779 | error = security_path_truncate(path); |
7715b521 AV |
2780 | if (!error) { |
2781 | error = do_truncate(path->dentry, 0, | |
2782 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2783 | filp); |
7715b521 AV |
2784 | } |
2785 | put_write_access(inode); | |
acd0c935 | 2786 | return error; |
1da177e4 LT |
2787 | } |
2788 | ||
d57999e1 DH |
2789 | static inline int open_to_namei_flags(int flag) |
2790 | { | |
8a5e929d AV |
2791 | if ((flag & O_ACCMODE) == 3) |
2792 | flag--; | |
d57999e1 DH |
2793 | return flag; |
2794 | } | |
2795 | ||
d18e9008 MS |
2796 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2797 | { | |
2798 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2799 | if (error) | |
2800 | return error; | |
2801 | ||
2802 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2803 | if (error) | |
2804 | return error; | |
2805 | ||
2806 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2807 | } | |
2808 | ||
1acf0af9 DH |
2809 | /* |
2810 | * Attempt to atomically look up, create and open a file from a negative | |
2811 | * dentry. | |
2812 | * | |
2813 | * Returns 0 if successful. The file will have been created and attached to | |
2814 | * @file by the filesystem calling finish_open(). | |
2815 | * | |
2816 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2817 | * caller will need to perform the open themselves. @path will have been | |
2818 | * updated to point to the new dentry. This may be negative. | |
2819 | * | |
2820 | * Returns an error code otherwise. | |
2821 | */ | |
2675a4eb AV |
2822 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2823 | struct path *path, struct file *file, | |
2824 | const struct open_flags *op, | |
64894cf8 | 2825 | bool got_write, bool need_lookup, |
2675a4eb | 2826 | int *opened) |
d18e9008 MS |
2827 | { |
2828 | struct inode *dir = nd->path.dentry->d_inode; | |
2829 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2830 | umode_t mode; | |
2831 | int error; | |
2832 | int acc_mode; | |
d18e9008 MS |
2833 | int create_error = 0; |
2834 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
116cc022 | 2835 | bool excl; |
d18e9008 MS |
2836 | |
2837 | BUG_ON(dentry->d_inode); | |
2838 | ||
2839 | /* Don't create child dentry for a dead directory. */ | |
2840 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2841 | error = -ENOENT; |
d18e9008 MS |
2842 | goto out; |
2843 | } | |
2844 | ||
62b259d8 | 2845 | mode = op->mode; |
d18e9008 MS |
2846 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2847 | mode &= ~current_umask(); | |
2848 | ||
116cc022 MS |
2849 | excl = (open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT); |
2850 | if (excl) | |
d18e9008 | 2851 | open_flag &= ~O_TRUNC; |
d18e9008 MS |
2852 | |
2853 | /* | |
2854 | * Checking write permission is tricky, bacuse we don't know if we are | |
2855 | * going to actually need it: O_CREAT opens should work as long as the | |
2856 | * file exists. But checking existence breaks atomicity. The trick is | |
2857 | * to check access and if not granted clear O_CREAT from the flags. | |
2858 | * | |
2859 | * Another problem is returing the "right" error value (e.g. for an | |
2860 | * O_EXCL open we want to return EEXIST not EROFS). | |
2861 | */ | |
64894cf8 AV |
2862 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2863 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2864 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2865 | /* |
2866 | * No O_CREATE -> atomicity not a requirement -> fall | |
2867 | * back to lookup + open | |
2868 | */ | |
2869 | goto no_open; | |
2870 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2871 | /* Fall back and fail with the right error */ | |
64894cf8 | 2872 | create_error = -EROFS; |
d18e9008 MS |
2873 | goto no_open; |
2874 | } else { | |
2875 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2876 | create_error = -EROFS; |
d18e9008 MS |
2877 | open_flag &= ~O_CREAT; |
2878 | } | |
2879 | } | |
2880 | ||
2881 | if (open_flag & O_CREAT) { | |
38227f78 | 2882 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2883 | if (error) { |
2884 | create_error = error; | |
2885 | if (open_flag & O_EXCL) | |
2886 | goto no_open; | |
2887 | open_flag &= ~O_CREAT; | |
2888 | } | |
2889 | } | |
2890 | ||
2891 | if (nd->flags & LOOKUP_DIRECTORY) | |
2892 | open_flag |= O_DIRECTORY; | |
2893 | ||
30d90494 AV |
2894 | file->f_path.dentry = DENTRY_NOT_SET; |
2895 | file->f_path.mnt = nd->path.mnt; | |
2896 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2897 | opened); |
d9585277 | 2898 | if (error < 0) { |
d9585277 AV |
2899 | if (create_error && error == -ENOENT) |
2900 | error = create_error; | |
d18e9008 MS |
2901 | goto out; |
2902 | } | |
2903 | ||
d9585277 | 2904 | if (error) { /* returned 1, that is */ |
30d90494 | 2905 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2906 | error = -EIO; |
d18e9008 MS |
2907 | goto out; |
2908 | } | |
30d90494 | 2909 | if (file->f_path.dentry) { |
d18e9008 | 2910 | dput(dentry); |
30d90494 | 2911 | dentry = file->f_path.dentry; |
d18e9008 | 2912 | } |
03da633a AV |
2913 | if (*opened & FILE_CREATED) |
2914 | fsnotify_create(dir, dentry); | |
2915 | if (!dentry->d_inode) { | |
2916 | WARN_ON(*opened & FILE_CREATED); | |
2917 | if (create_error) { | |
2918 | error = create_error; | |
2919 | goto out; | |
2920 | } | |
2921 | } else { | |
2922 | if (excl && !(*opened & FILE_CREATED)) { | |
2923 | error = -EEXIST; | |
2924 | goto out; | |
2925 | } | |
62b2ce96 | 2926 | } |
d18e9008 MS |
2927 | goto looked_up; |
2928 | } | |
2929 | ||
2930 | /* | |
2931 | * We didn't have the inode before the open, so check open permission | |
2932 | * here. | |
2933 | */ | |
03da633a AV |
2934 | acc_mode = op->acc_mode; |
2935 | if (*opened & FILE_CREATED) { | |
2936 | WARN_ON(!(open_flag & O_CREAT)); | |
2937 | fsnotify_create(dir, dentry); | |
62fb4a15 | 2938 | acc_mode = 0; |
03da633a | 2939 | } |
2675a4eb AV |
2940 | error = may_open(&file->f_path, acc_mode, open_flag); |
2941 | if (error) | |
2942 | fput(file); | |
d18e9008 MS |
2943 | |
2944 | out: | |
2945 | dput(dentry); | |
2675a4eb | 2946 | return error; |
d18e9008 | 2947 | |
d18e9008 MS |
2948 | no_open: |
2949 | if (need_lookup) { | |
72bd866a | 2950 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2951 | if (IS_ERR(dentry)) |
2675a4eb | 2952 | return PTR_ERR(dentry); |
d18e9008 MS |
2953 | |
2954 | if (create_error) { | |
2955 | int open_flag = op->open_flag; | |
2956 | ||
2675a4eb | 2957 | error = create_error; |
d18e9008 MS |
2958 | if ((open_flag & O_EXCL)) { |
2959 | if (!dentry->d_inode) | |
2960 | goto out; | |
2961 | } else if (!dentry->d_inode) { | |
2962 | goto out; | |
2963 | } else if ((open_flag & O_TRUNC) && | |
e36cb0b8 | 2964 | d_is_reg(dentry)) { |
d18e9008 MS |
2965 | goto out; |
2966 | } | |
2967 | /* will fail later, go on to get the right error */ | |
2968 | } | |
2969 | } | |
2970 | looked_up: | |
2971 | path->dentry = dentry; | |
2972 | path->mnt = nd->path.mnt; | |
2675a4eb | 2973 | return 1; |
d18e9008 MS |
2974 | } |
2975 | ||
d58ffd35 | 2976 | /* |
1acf0af9 | 2977 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2978 | * |
2979 | * Must be called with i_mutex held on parent. | |
2980 | * | |
1acf0af9 DH |
2981 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2982 | * opened. In this case the file will be returned attached to @file. | |
2983 | * | |
2984 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2985 | * and creations will have been performed and the dentry returned in @path will | |
2986 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2987 | * specified then a negative dentry may be returned. | |
2988 | * | |
2989 | * An error code is returned otherwise. | |
2990 | * | |
2991 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2992 | * cleared otherwise prior to returning. | |
d58ffd35 | 2993 | */ |
2675a4eb AV |
2994 | static int lookup_open(struct nameidata *nd, struct path *path, |
2995 | struct file *file, | |
2996 | const struct open_flags *op, | |
64894cf8 | 2997 | bool got_write, int *opened) |
d58ffd35 MS |
2998 | { |
2999 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 3000 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
3001 | struct dentry *dentry; |
3002 | int error; | |
6c51e513 | 3003 | bool need_lookup = false; |
d58ffd35 | 3004 | |
47237687 | 3005 | *opened &= ~FILE_CREATED; |
6c51e513 | 3006 | dentry = lookup_dcache(&nd->last, dir, nd->flags); |
d58ffd35 | 3007 | if (IS_ERR(dentry)) |
2675a4eb | 3008 | return PTR_ERR(dentry); |
d58ffd35 | 3009 | |
6c51e513 AV |
3010 | if (!dentry) { |
3011 | dentry = d_alloc(dir, &nd->last); | |
3012 | if (unlikely(!dentry)) | |
3013 | return -ENOMEM; | |
3014 | need_lookup = true; | |
3015 | } else if (dentry->d_inode) { | |
3016 | /* Cached positive dentry: will open in f_op->open */ | |
d18e9008 | 3017 | goto out_no_open; |
6c51e513 | 3018 | } |
d18e9008 MS |
3019 | |
3020 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 3021 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 3022 | need_lookup, opened); |
d18e9008 MS |
3023 | } |
3024 | ||
54ef4872 MS |
3025 | if (need_lookup) { |
3026 | BUG_ON(dentry->d_inode); | |
3027 | ||
72bd866a | 3028 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 3029 | if (IS_ERR(dentry)) |
2675a4eb | 3030 | return PTR_ERR(dentry); |
54ef4872 MS |
3031 | } |
3032 | ||
d58ffd35 MS |
3033 | /* Negative dentry, just create the file */ |
3034 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
3035 | umode_t mode = op->mode; | |
3036 | if (!IS_POSIXACL(dir->d_inode)) | |
3037 | mode &= ~current_umask(); | |
3038 | /* | |
3039 | * This write is needed to ensure that a | |
3040 | * rw->ro transition does not occur between | |
3041 | * the time when the file is created and when | |
3042 | * a permanent write count is taken through | |
015c3bbc | 3043 | * the 'struct file' in finish_open(). |
d58ffd35 | 3044 | */ |
64894cf8 AV |
3045 | if (!got_write) { |
3046 | error = -EROFS; | |
d58ffd35 | 3047 | goto out_dput; |
64894cf8 | 3048 | } |
47237687 | 3049 | *opened |= FILE_CREATED; |
d58ffd35 MS |
3050 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
3051 | if (error) | |
3052 | goto out_dput; | |
312b63fb AV |
3053 | error = vfs_create(dir->d_inode, dentry, mode, |
3054 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
3055 | if (error) |
3056 | goto out_dput; | |
3057 | } | |
d18e9008 | 3058 | out_no_open: |
d58ffd35 MS |
3059 | path->dentry = dentry; |
3060 | path->mnt = nd->path.mnt; | |
2675a4eb | 3061 | return 1; |
d58ffd35 MS |
3062 | |
3063 | out_dput: | |
3064 | dput(dentry); | |
2675a4eb | 3065 | return error; |
d58ffd35 MS |
3066 | } |
3067 | ||
31e6b01f | 3068 | /* |
fe2d35ff | 3069 | * Handle the last step of open() |
31e6b01f | 3070 | */ |
896475d5 | 3071 | static int do_last(struct nameidata *nd, |
2675a4eb | 3072 | struct file *file, const struct open_flags *op, |
76ae2a5a | 3073 | int *opened) |
fb1cc555 | 3074 | { |
a1e28038 | 3075 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 3076 | int open_flag = op->open_flag; |
77d660a8 | 3077 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 3078 | bool got_write = false; |
bcda7652 | 3079 | int acc_mode = op->acc_mode; |
254cf582 | 3080 | unsigned seq; |
a1eb3315 | 3081 | struct inode *inode; |
16b1c1cd | 3082 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
896475d5 | 3083 | struct path path; |
16b1c1cd | 3084 | bool retried = false; |
16c2cd71 | 3085 | int error; |
1f36f774 | 3086 | |
c3e380b0 AV |
3087 | nd->flags &= ~LOOKUP_PARENT; |
3088 | nd->flags |= op->intent; | |
3089 | ||
bc77daa7 | 3090 | if (nd->last_type != LAST_NORM) { |
fe2d35ff | 3091 | error = handle_dots(nd, nd->last_type); |
deb106c6 | 3092 | if (unlikely(error)) |
2675a4eb | 3093 | return error; |
e83db167 | 3094 | goto finish_open; |
1f36f774 | 3095 | } |
67ee3ad2 | 3096 | |
ca344a89 | 3097 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
3098 | if (nd->last.name[nd->last.len]) |
3099 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
3100 | /* we _can_ be in RCU mode here */ | |
254cf582 | 3101 | error = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 3102 | if (likely(error > 0)) |
71574865 MS |
3103 | goto finish_lookup; |
3104 | ||
3105 | if (error < 0) | |
deb106c6 | 3106 | return error; |
71574865 MS |
3107 | |
3108 | BUG_ON(nd->inode != dir->d_inode); | |
6583fe22 | 3109 | BUG_ON(nd->flags & LOOKUP_RCU); |
b6183df7 MS |
3110 | } else { |
3111 | /* create side of things */ | |
3112 | /* | |
3113 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
3114 | * has been cleared when we got to the last component we are | |
3115 | * about to look up | |
3116 | */ | |
3117 | error = complete_walk(nd); | |
e8bb73df | 3118 | if (error) |
2675a4eb | 3119 | return error; |
fe2d35ff | 3120 | |
76ae2a5a | 3121 | audit_inode(nd->name, dir, LOOKUP_PARENT); |
b6183df7 | 3122 | /* trailing slashes? */ |
deb106c6 AV |
3123 | if (unlikely(nd->last.name[nd->last.len])) |
3124 | return -EISDIR; | |
b6183df7 | 3125 | } |
a2c36b45 | 3126 | |
16b1c1cd | 3127 | retry_lookup: |
64894cf8 AV |
3128 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
3129 | error = mnt_want_write(nd->path.mnt); | |
3130 | if (!error) | |
3131 | got_write = true; | |
3132 | /* | |
3133 | * do _not_ fail yet - we might not need that or fail with | |
3134 | * a different error; let lookup_open() decide; we'll be | |
3135 | * dropping this one anyway. | |
3136 | */ | |
3137 | } | |
5955102c | 3138 | inode_lock(dir->d_inode); |
896475d5 | 3139 | error = lookup_open(nd, &path, file, op, got_write, opened); |
5955102c | 3140 | inode_unlock(dir->d_inode); |
a1e28038 | 3141 | |
2675a4eb AV |
3142 | if (error <= 0) { |
3143 | if (error) | |
d18e9008 MS |
3144 | goto out; |
3145 | ||
47237687 | 3146 | if ((*opened & FILE_CREATED) || |
496ad9aa | 3147 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3148 | will_truncate = false; |
d18e9008 | 3149 | |
76ae2a5a | 3150 | audit_inode(nd->name, file->f_path.dentry, 0); |
d18e9008 MS |
3151 | goto opened; |
3152 | } | |
fb1cc555 | 3153 | |
47237687 | 3154 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 3155 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3156 | open_flag &= ~O_TRUNC; |
77d660a8 | 3157 | will_truncate = false; |
62fb4a15 | 3158 | acc_mode = 0; |
896475d5 | 3159 | path_to_nameidata(&path, nd); |
e83db167 | 3160 | goto finish_open_created; |
fb1cc555 AV |
3161 | } |
3162 | ||
d18e9008 MS |
3163 | /* |
3164 | * If atomic_open() acquired write access it is dropped now due to | |
3165 | * possible mount and symlink following (this might be optimized away if | |
3166 | * necessary...) | |
3167 | */ | |
64894cf8 | 3168 | if (got_write) { |
d18e9008 | 3169 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3170 | got_write = false; |
d18e9008 MS |
3171 | } |
3172 | ||
6583fe22 AV |
3173 | if (unlikely(d_is_negative(path.dentry))) { |
3174 | path_to_nameidata(&path, nd); | |
3175 | return -ENOENT; | |
3176 | } | |
3177 | ||
3178 | /* | |
3179 | * create/update audit record if it already exists. | |
3180 | */ | |
3181 | audit_inode(nd->name, path.dentry, 0); | |
3182 | ||
deb106c6 AV |
3183 | if (unlikely((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT))) { |
3184 | path_to_nameidata(&path, nd); | |
3185 | return -EEXIST; | |
3186 | } | |
fb1cc555 | 3187 | |
756daf26 | 3188 | error = follow_managed(&path, nd); |
deb106c6 AV |
3189 | if (unlikely(error < 0)) |
3190 | return error; | |
a3fbbde7 | 3191 | |
254cf582 | 3192 | seq = 0; /* out of RCU mode, so the value doesn't matter */ |
d4565649 | 3193 | inode = d_backing_inode(path.dentry); |
766c4cbf | 3194 | finish_lookup: |
d63ff28f AV |
3195 | if (nd->depth) |
3196 | put_link(nd); | |
181548c0 AV |
3197 | error = should_follow_link(nd, &path, nd->flags & LOOKUP_FOLLOW, |
3198 | inode, seq); | |
deb106c6 | 3199 | if (unlikely(error)) |
d63ff28f | 3200 | return error; |
fb1cc555 | 3201 | |
896475d5 AV |
3202 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path.mnt) { |
3203 | path_to_nameidata(&path, nd); | |
16b1c1cd MS |
3204 | } else { |
3205 | save_parent.dentry = nd->path.dentry; | |
896475d5 AV |
3206 | save_parent.mnt = mntget(path.mnt); |
3207 | nd->path.dentry = path.dentry; | |
16b1c1cd MS |
3208 | |
3209 | } | |
decf3400 | 3210 | nd->inode = inode; |
254cf582 | 3211 | nd->seq = seq; |
a3fbbde7 | 3212 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
bc77daa7 | 3213 | finish_open: |
a3fbbde7 | 3214 | error = complete_walk(nd); |
16b1c1cd MS |
3215 | if (error) { |
3216 | path_put(&save_parent); | |
2675a4eb | 3217 | return error; |
16b1c1cd | 3218 | } |
76ae2a5a | 3219 | audit_inode(nd->name, nd->path.dentry, 0); |
5129fa48 AV |
3220 | if (unlikely(d_is_symlink(nd->path.dentry)) && !(open_flag & O_PATH)) { |
3221 | error = -ELOOP; | |
3222 | goto out; | |
3223 | } | |
fb1cc555 | 3224 | error = -EISDIR; |
44b1d530 | 3225 | if ((open_flag & O_CREAT) && d_is_dir(nd->path.dentry)) |
2675a4eb | 3226 | goto out; |
af2f5542 | 3227 | error = -ENOTDIR; |
44b1d530 | 3228 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
2675a4eb | 3229 | goto out; |
4bbcbd3b | 3230 | if (!d_is_reg(nd->path.dentry)) |
77d660a8 | 3231 | will_truncate = false; |
6c0d46c4 | 3232 | |
0f9d1a10 AV |
3233 | if (will_truncate) { |
3234 | error = mnt_want_write(nd->path.mnt); | |
3235 | if (error) | |
2675a4eb | 3236 | goto out; |
64894cf8 | 3237 | got_write = true; |
0f9d1a10 | 3238 | } |
e83db167 | 3239 | finish_open_created: |
62fb4a15 AV |
3240 | if (likely(!(open_flag & O_PATH))) { |
3241 | error = may_open(&nd->path, acc_mode, open_flag); | |
3242 | if (error) | |
3243 | goto out; | |
3244 | } | |
4aa7c634 MS |
3245 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
3246 | error = vfs_open(&nd->path, file, current_cred()); | |
3247 | if (!error) { | |
3248 | *opened |= FILE_OPENED; | |
3249 | } else { | |
30d90494 | 3250 | if (error == -EOPENSTALE) |
f60dc3db | 3251 | goto stale_open; |
015c3bbc | 3252 | goto out; |
f60dc3db | 3253 | } |
a8277b9b | 3254 | opened: |
2675a4eb | 3255 | error = open_check_o_direct(file); |
015c3bbc MS |
3256 | if (error) |
3257 | goto exit_fput; | |
3034a146 | 3258 | error = ima_file_check(file, op->acc_mode, *opened); |
aa4caadb MS |
3259 | if (error) |
3260 | goto exit_fput; | |
3261 | ||
3262 | if (will_truncate) { | |
2675a4eb | 3263 | error = handle_truncate(file); |
aa4caadb MS |
3264 | if (error) |
3265 | goto exit_fput; | |
0f9d1a10 | 3266 | } |
ca344a89 | 3267 | out: |
c80567c8 AV |
3268 | if (unlikely(error > 0)) { |
3269 | WARN_ON(1); | |
3270 | error = -EINVAL; | |
3271 | } | |
64894cf8 | 3272 | if (got_write) |
0f9d1a10 | 3273 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 3274 | path_put(&save_parent); |
2675a4eb | 3275 | return error; |
fb1cc555 | 3276 | |
015c3bbc | 3277 | exit_fput: |
2675a4eb AV |
3278 | fput(file); |
3279 | goto out; | |
015c3bbc | 3280 | |
f60dc3db MS |
3281 | stale_open: |
3282 | /* If no saved parent or already retried then can't retry */ | |
3283 | if (!save_parent.dentry || retried) | |
3284 | goto out; | |
3285 | ||
3286 | BUG_ON(save_parent.dentry != dir); | |
3287 | path_put(&nd->path); | |
3288 | nd->path = save_parent; | |
3289 | nd->inode = dir->d_inode; | |
3290 | save_parent.mnt = NULL; | |
3291 | save_parent.dentry = NULL; | |
64894cf8 | 3292 | if (got_write) { |
f60dc3db | 3293 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3294 | got_write = false; |
f60dc3db MS |
3295 | } |
3296 | retried = true; | |
3297 | goto retry_lookup; | |
fb1cc555 AV |
3298 | } |
3299 | ||
c8a53ee5 | 3300 | static int do_tmpfile(struct nameidata *nd, unsigned flags, |
60545d0d AV |
3301 | const struct open_flags *op, |
3302 | struct file *file, int *opened) | |
3303 | { | |
3304 | static const struct qstr name = QSTR_INIT("/", 1); | |
625b6d10 | 3305 | struct dentry *child; |
60545d0d | 3306 | struct inode *dir; |
625b6d10 | 3307 | struct path path; |
c8a53ee5 | 3308 | int error = path_lookupat(nd, flags | LOOKUP_DIRECTORY, &path); |
60545d0d AV |
3309 | if (unlikely(error)) |
3310 | return error; | |
625b6d10 | 3311 | error = mnt_want_write(path.mnt); |
60545d0d AV |
3312 | if (unlikely(error)) |
3313 | goto out; | |
625b6d10 | 3314 | dir = path.dentry->d_inode; |
60545d0d | 3315 | /* we want directory to be writable */ |
625b6d10 | 3316 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
60545d0d AV |
3317 | if (error) |
3318 | goto out2; | |
60545d0d AV |
3319 | if (!dir->i_op->tmpfile) { |
3320 | error = -EOPNOTSUPP; | |
3321 | goto out2; | |
3322 | } | |
625b6d10 | 3323 | child = d_alloc(path.dentry, &name); |
60545d0d AV |
3324 | if (unlikely(!child)) { |
3325 | error = -ENOMEM; | |
3326 | goto out2; | |
3327 | } | |
625b6d10 AV |
3328 | dput(path.dentry); |
3329 | path.dentry = child; | |
3330 | error = dir->i_op->tmpfile(dir, child, op->mode); | |
60545d0d AV |
3331 | if (error) |
3332 | goto out2; | |
c8a53ee5 | 3333 | audit_inode(nd->name, child, 0); |
69a91c23 | 3334 | /* Don't check for other permissions, the inode was just created */ |
62fb4a15 | 3335 | error = may_open(&path, 0, op->open_flag); |
60545d0d AV |
3336 | if (error) |
3337 | goto out2; | |
625b6d10 AV |
3338 | file->f_path.mnt = path.mnt; |
3339 | error = finish_open(file, child, NULL, opened); | |
60545d0d AV |
3340 | if (error) |
3341 | goto out2; | |
3342 | error = open_check_o_direct(file); | |
f4e0c30c | 3343 | if (error) { |
60545d0d | 3344 | fput(file); |
f4e0c30c AV |
3345 | } else if (!(op->open_flag & O_EXCL)) { |
3346 | struct inode *inode = file_inode(file); | |
3347 | spin_lock(&inode->i_lock); | |
3348 | inode->i_state |= I_LINKABLE; | |
3349 | spin_unlock(&inode->i_lock); | |
3350 | } | |
60545d0d | 3351 | out2: |
625b6d10 | 3352 | mnt_drop_write(path.mnt); |
60545d0d | 3353 | out: |
625b6d10 | 3354 | path_put(&path); |
60545d0d AV |
3355 | return error; |
3356 | } | |
3357 | ||
c8a53ee5 AV |
3358 | static struct file *path_openat(struct nameidata *nd, |
3359 | const struct open_flags *op, unsigned flags) | |
1da177e4 | 3360 | { |
368ee9ba | 3361 | const char *s; |
30d90494 | 3362 | struct file *file; |
47237687 | 3363 | int opened = 0; |
13aab428 | 3364 | int error; |
31e6b01f | 3365 | |
30d90494 | 3366 | file = get_empty_filp(); |
1afc99be AV |
3367 | if (IS_ERR(file)) |
3368 | return file; | |
31e6b01f | 3369 | |
30d90494 | 3370 | file->f_flags = op->open_flag; |
31e6b01f | 3371 | |
bb458c64 | 3372 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
c8a53ee5 | 3373 | error = do_tmpfile(nd, flags, op, file, &opened); |
f15133df | 3374 | goto out2; |
60545d0d AV |
3375 | } |
3376 | ||
c8a53ee5 | 3377 | s = path_init(nd, flags); |
368ee9ba AV |
3378 | if (IS_ERR(s)) { |
3379 | put_filp(file); | |
3380 | return ERR_CAST(s); | |
3381 | } | |
3bdba28b | 3382 | while (!(error = link_path_walk(s, nd)) && |
76ae2a5a | 3383 | (error = do_last(nd, file, op, &opened)) > 0) { |
73d049a4 | 3384 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); |
3bdba28b AV |
3385 | s = trailing_symlink(nd); |
3386 | if (IS_ERR(s)) { | |
3387 | error = PTR_ERR(s); | |
2675a4eb | 3388 | break; |
3bdba28b | 3389 | } |
806b681c | 3390 | } |
deb106c6 | 3391 | terminate_walk(nd); |
f15133df | 3392 | out2: |
2675a4eb AV |
3393 | if (!(opened & FILE_OPENED)) { |
3394 | BUG_ON(!error); | |
30d90494 | 3395 | put_filp(file); |
16b1c1cd | 3396 | } |
2675a4eb AV |
3397 | if (unlikely(error)) { |
3398 | if (error == -EOPENSTALE) { | |
3399 | if (flags & LOOKUP_RCU) | |
3400 | error = -ECHILD; | |
3401 | else | |
3402 | error = -ESTALE; | |
3403 | } | |
3404 | file = ERR_PTR(error); | |
3405 | } | |
3406 | return file; | |
1da177e4 LT |
3407 | } |
3408 | ||
669abf4e | 3409 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3410 | const struct open_flags *op) |
13aab428 | 3411 | { |
9883d185 | 3412 | struct nameidata nd; |
f9652e10 | 3413 | int flags = op->lookup_flags; |
13aab428 AV |
3414 | struct file *filp; |
3415 | ||
9883d185 | 3416 | set_nameidata(&nd, dfd, pathname); |
c8a53ee5 | 3417 | filp = path_openat(&nd, op, flags | LOOKUP_RCU); |
13aab428 | 3418 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3419 | filp = path_openat(&nd, op, flags); |
13aab428 | 3420 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3421 | filp = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3422 | restore_nameidata(); |
13aab428 AV |
3423 | return filp; |
3424 | } | |
3425 | ||
73d049a4 | 3426 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3427 | const char *name, const struct open_flags *op) |
73d049a4 | 3428 | { |
9883d185 | 3429 | struct nameidata nd; |
73d049a4 | 3430 | struct file *file; |
51689104 | 3431 | struct filename *filename; |
f9652e10 | 3432 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3433 | |
3434 | nd.root.mnt = mnt; | |
3435 | nd.root.dentry = dentry; | |
3436 | ||
b18825a7 | 3437 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3438 | return ERR_PTR(-ELOOP); |
3439 | ||
51689104 | 3440 | filename = getname_kernel(name); |
a1c83681 | 3441 | if (IS_ERR(filename)) |
51689104 PM |
3442 | return ERR_CAST(filename); |
3443 | ||
9883d185 | 3444 | set_nameidata(&nd, -1, filename); |
c8a53ee5 | 3445 | file = path_openat(&nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3446 | if (unlikely(file == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3447 | file = path_openat(&nd, op, flags); |
73d049a4 | 3448 | if (unlikely(file == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3449 | file = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3450 | restore_nameidata(); |
51689104 | 3451 | putname(filename); |
73d049a4 AV |
3452 | return file; |
3453 | } | |
3454 | ||
fa14a0b8 | 3455 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3456 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3457 | { |
c663e5d8 | 3458 | struct dentry *dentry = ERR_PTR(-EEXIST); |
391172c4 AV |
3459 | struct qstr last; |
3460 | int type; | |
c30dabfe | 3461 | int err2; |
1ac12b4b JL |
3462 | int error; |
3463 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3464 | ||
3465 | /* | |
3466 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3467 | * other flags passed in are ignored! | |
3468 | */ | |
3469 | lookup_flags &= LOOKUP_REVAL; | |
3470 | ||
5c31b6ce AV |
3471 | name = filename_parentat(dfd, name, lookup_flags, path, &last, &type); |
3472 | if (IS_ERR(name)) | |
3473 | return ERR_CAST(name); | |
1da177e4 | 3474 | |
c663e5d8 CH |
3475 | /* |
3476 | * Yucky last component or no last component at all? | |
3477 | * (foo/., foo/.., /////) | |
3478 | */ | |
5c31b6ce | 3479 | if (unlikely(type != LAST_NORM)) |
ed75e95d | 3480 | goto out; |
c663e5d8 | 3481 | |
c30dabfe | 3482 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
391172c4 | 3483 | err2 = mnt_want_write(path->mnt); |
c663e5d8 CH |
3484 | /* |
3485 | * Do the final lookup. | |
3486 | */ | |
391172c4 | 3487 | lookup_flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
5955102c | 3488 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 3489 | dentry = __lookup_hash(&last, path->dentry, lookup_flags); |
1da177e4 | 3490 | if (IS_ERR(dentry)) |
a8104a9f | 3491 | goto unlock; |
c663e5d8 | 3492 | |
a8104a9f | 3493 | error = -EEXIST; |
b18825a7 | 3494 | if (d_is_positive(dentry)) |
a8104a9f | 3495 | goto fail; |
b18825a7 | 3496 | |
c663e5d8 CH |
3497 | /* |
3498 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3499 | * From the vfs_mknod() POV we just have a negative dentry - | |
3500 | * all is fine. Let's be bastards - you had / on the end, you've | |
3501 | * been asking for (non-existent) directory. -ENOENT for you. | |
3502 | */ | |
391172c4 | 3503 | if (unlikely(!is_dir && last.name[last.len])) { |
a8104a9f | 3504 | error = -ENOENT; |
ed75e95d | 3505 | goto fail; |
e9baf6e5 | 3506 | } |
c30dabfe JK |
3507 | if (unlikely(err2)) { |
3508 | error = err2; | |
a8104a9f | 3509 | goto fail; |
c30dabfe | 3510 | } |
181c37b6 | 3511 | putname(name); |
1da177e4 | 3512 | return dentry; |
1da177e4 | 3513 | fail: |
a8104a9f AV |
3514 | dput(dentry); |
3515 | dentry = ERR_PTR(error); | |
3516 | unlock: | |
5955102c | 3517 | inode_unlock(path->dentry->d_inode); |
c30dabfe | 3518 | if (!err2) |
391172c4 | 3519 | mnt_drop_write(path->mnt); |
ed75e95d | 3520 | out: |
391172c4 | 3521 | path_put(path); |
181c37b6 | 3522 | putname(name); |
1da177e4 LT |
3523 | return dentry; |
3524 | } | |
fa14a0b8 AV |
3525 | |
3526 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3527 | struct path *path, unsigned int lookup_flags) | |
3528 | { | |
181c37b6 AV |
3529 | return filename_create(dfd, getname_kernel(pathname), |
3530 | path, lookup_flags); | |
fa14a0b8 | 3531 | } |
dae6ad8f AV |
3532 | EXPORT_SYMBOL(kern_path_create); |
3533 | ||
921a1650 AV |
3534 | void done_path_create(struct path *path, struct dentry *dentry) |
3535 | { | |
3536 | dput(dentry); | |
5955102c | 3537 | inode_unlock(path->dentry->d_inode); |
a8104a9f | 3538 | mnt_drop_write(path->mnt); |
921a1650 AV |
3539 | path_put(path); |
3540 | } | |
3541 | EXPORT_SYMBOL(done_path_create); | |
3542 | ||
520ae687 | 3543 | inline struct dentry *user_path_create(int dfd, const char __user *pathname, |
1ac12b4b | 3544 | struct path *path, unsigned int lookup_flags) |
dae6ad8f | 3545 | { |
181c37b6 | 3546 | return filename_create(dfd, getname(pathname), path, lookup_flags); |
dae6ad8f AV |
3547 | } |
3548 | EXPORT_SYMBOL(user_path_create); | |
3549 | ||
1a67aafb | 3550 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3551 | { |
a95164d9 | 3552 | int error = may_create(dir, dentry); |
1da177e4 LT |
3553 | |
3554 | if (error) | |
3555 | return error; | |
3556 | ||
975d6b39 | 3557 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3558 | return -EPERM; |
3559 | ||
acfa4380 | 3560 | if (!dir->i_op->mknod) |
1da177e4 LT |
3561 | return -EPERM; |
3562 | ||
08ce5f16 SH |
3563 | error = devcgroup_inode_mknod(mode, dev); |
3564 | if (error) | |
3565 | return error; | |
3566 | ||
1da177e4 LT |
3567 | error = security_inode_mknod(dir, dentry, mode, dev); |
3568 | if (error) | |
3569 | return error; | |
3570 | ||
1da177e4 | 3571 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3572 | if (!error) |
f38aa942 | 3573 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3574 | return error; |
3575 | } | |
4d359507 | 3576 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3577 | |
f69aac00 | 3578 | static int may_mknod(umode_t mode) |
463c3197 DH |
3579 | { |
3580 | switch (mode & S_IFMT) { | |
3581 | case S_IFREG: | |
3582 | case S_IFCHR: | |
3583 | case S_IFBLK: | |
3584 | case S_IFIFO: | |
3585 | case S_IFSOCK: | |
3586 | case 0: /* zero mode translates to S_IFREG */ | |
3587 | return 0; | |
3588 | case S_IFDIR: | |
3589 | return -EPERM; | |
3590 | default: | |
3591 | return -EINVAL; | |
3592 | } | |
3593 | } | |
3594 | ||
8208a22b | 3595 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3596 | unsigned, dev) |
1da177e4 | 3597 | { |
2ad94ae6 | 3598 | struct dentry *dentry; |
dae6ad8f AV |
3599 | struct path path; |
3600 | int error; | |
972567f1 | 3601 | unsigned int lookup_flags = 0; |
1da177e4 | 3602 | |
8e4bfca1 AV |
3603 | error = may_mknod(mode); |
3604 | if (error) | |
3605 | return error; | |
972567f1 JL |
3606 | retry: |
3607 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3608 | if (IS_ERR(dentry)) |
3609 | return PTR_ERR(dentry); | |
2ad94ae6 | 3610 | |
dae6ad8f | 3611 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3612 | mode &= ~current_umask(); |
dae6ad8f | 3613 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3614 | if (error) |
a8104a9f | 3615 | goto out; |
463c3197 | 3616 | switch (mode & S_IFMT) { |
1da177e4 | 3617 | case 0: case S_IFREG: |
312b63fb | 3618 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3619 | break; |
3620 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3621 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3622 | new_decode_dev(dev)); |
3623 | break; | |
3624 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3625 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3626 | break; |
1da177e4 | 3627 | } |
a8104a9f | 3628 | out: |
921a1650 | 3629 | done_path_create(&path, dentry); |
972567f1 JL |
3630 | if (retry_estale(error, lookup_flags)) { |
3631 | lookup_flags |= LOOKUP_REVAL; | |
3632 | goto retry; | |
3633 | } | |
1da177e4 LT |
3634 | return error; |
3635 | } | |
3636 | ||
8208a22b | 3637 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3638 | { |
3639 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3640 | } | |
3641 | ||
18bb1db3 | 3642 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3643 | { |
a95164d9 | 3644 | int error = may_create(dir, dentry); |
8de52778 | 3645 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3646 | |
3647 | if (error) | |
3648 | return error; | |
3649 | ||
acfa4380 | 3650 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3651 | return -EPERM; |
3652 | ||
3653 | mode &= (S_IRWXUGO|S_ISVTX); | |
3654 | error = security_inode_mkdir(dir, dentry, mode); | |
3655 | if (error) | |
3656 | return error; | |
3657 | ||
8de52778 AV |
3658 | if (max_links && dir->i_nlink >= max_links) |
3659 | return -EMLINK; | |
3660 | ||
1da177e4 | 3661 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3662 | if (!error) |
f38aa942 | 3663 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3664 | return error; |
3665 | } | |
4d359507 | 3666 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3667 | |
a218d0fd | 3668 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3669 | { |
6902d925 | 3670 | struct dentry *dentry; |
dae6ad8f AV |
3671 | struct path path; |
3672 | int error; | |
b76d8b82 | 3673 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3674 | |
b76d8b82 JL |
3675 | retry: |
3676 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3677 | if (IS_ERR(dentry)) |
dae6ad8f | 3678 | return PTR_ERR(dentry); |
1da177e4 | 3679 | |
dae6ad8f | 3680 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3681 | mode &= ~current_umask(); |
dae6ad8f | 3682 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3683 | if (!error) |
3684 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3685 | done_path_create(&path, dentry); |
b76d8b82 JL |
3686 | if (retry_estale(error, lookup_flags)) { |
3687 | lookup_flags |= LOOKUP_REVAL; | |
3688 | goto retry; | |
3689 | } | |
1da177e4 LT |
3690 | return error; |
3691 | } | |
3692 | ||
a218d0fd | 3693 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3694 | { |
3695 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3696 | } | |
3697 | ||
1da177e4 | 3698 | /* |
a71905f0 | 3699 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3700 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3701 | * dentry, and if that is true (possibly after pruning the dcache), |
3702 | * then we drop the dentry now. | |
1da177e4 LT |
3703 | * |
3704 | * A low-level filesystem can, if it choses, legally | |
3705 | * do a | |
3706 | * | |
3707 | * if (!d_unhashed(dentry)) | |
3708 | * return -EBUSY; | |
3709 | * | |
3710 | * if it cannot handle the case of removing a directory | |
3711 | * that is still in use by something else.. | |
3712 | */ | |
3713 | void dentry_unhash(struct dentry *dentry) | |
3714 | { | |
dc168427 | 3715 | shrink_dcache_parent(dentry); |
1da177e4 | 3716 | spin_lock(&dentry->d_lock); |
98474236 | 3717 | if (dentry->d_lockref.count == 1) |
1da177e4 LT |
3718 | __d_drop(dentry); |
3719 | spin_unlock(&dentry->d_lock); | |
1da177e4 | 3720 | } |
4d359507 | 3721 | EXPORT_SYMBOL(dentry_unhash); |
1da177e4 LT |
3722 | |
3723 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3724 | { | |
3725 | int error = may_delete(dir, dentry, 1); | |
3726 | ||
3727 | if (error) | |
3728 | return error; | |
3729 | ||
acfa4380 | 3730 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3731 | return -EPERM; |
3732 | ||
1d2ef590 | 3733 | dget(dentry); |
5955102c | 3734 | inode_lock(dentry->d_inode); |
912dbc15 SW |
3735 | |
3736 | error = -EBUSY; | |
7af1364f | 3737 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3738 | goto out; |
3739 | ||
3740 | error = security_inode_rmdir(dir, dentry); | |
3741 | if (error) | |
3742 | goto out; | |
3743 | ||
3cebde24 | 3744 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3745 | error = dir->i_op->rmdir(dir, dentry); |
3746 | if (error) | |
3747 | goto out; | |
3748 | ||
3749 | dentry->d_inode->i_flags |= S_DEAD; | |
3750 | dont_mount(dentry); | |
8ed936b5 | 3751 | detach_mounts(dentry); |
912dbc15 SW |
3752 | |
3753 | out: | |
5955102c | 3754 | inode_unlock(dentry->d_inode); |
1d2ef590 | 3755 | dput(dentry); |
912dbc15 | 3756 | if (!error) |
1da177e4 | 3757 | d_delete(dentry); |
1da177e4 LT |
3758 | return error; |
3759 | } | |
4d359507 | 3760 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3761 | |
5590ff0d | 3762 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3763 | { |
3764 | int error = 0; | |
91a27b2a | 3765 | struct filename *name; |
1da177e4 | 3766 | struct dentry *dentry; |
f5beed75 AV |
3767 | struct path path; |
3768 | struct qstr last; | |
3769 | int type; | |
c6ee9206 JL |
3770 | unsigned int lookup_flags = 0; |
3771 | retry: | |
f5beed75 AV |
3772 | name = user_path_parent(dfd, pathname, |
3773 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3774 | if (IS_ERR(name)) |
3775 | return PTR_ERR(name); | |
1da177e4 | 3776 | |
f5beed75 | 3777 | switch (type) { |
0612d9fb OH |
3778 | case LAST_DOTDOT: |
3779 | error = -ENOTEMPTY; | |
3780 | goto exit1; | |
3781 | case LAST_DOT: | |
3782 | error = -EINVAL; | |
3783 | goto exit1; | |
3784 | case LAST_ROOT: | |
3785 | error = -EBUSY; | |
3786 | goto exit1; | |
1da177e4 | 3787 | } |
0612d9fb | 3788 | |
f5beed75 | 3789 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3790 | if (error) |
3791 | goto exit1; | |
0612d9fb | 3792 | |
5955102c | 3793 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3794 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 | 3795 | error = PTR_ERR(dentry); |
6902d925 DH |
3796 | if (IS_ERR(dentry)) |
3797 | goto exit2; | |
e6bc45d6 TT |
3798 | if (!dentry->d_inode) { |
3799 | error = -ENOENT; | |
3800 | goto exit3; | |
3801 | } | |
f5beed75 | 3802 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3803 | if (error) |
c30dabfe | 3804 | goto exit3; |
f5beed75 | 3805 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3806 | exit3: |
6902d925 DH |
3807 | dput(dentry); |
3808 | exit2: | |
5955102c | 3809 | inode_unlock(path.dentry->d_inode); |
f5beed75 | 3810 | mnt_drop_write(path.mnt); |
1da177e4 | 3811 | exit1: |
f5beed75 | 3812 | path_put(&path); |
1da177e4 | 3813 | putname(name); |
c6ee9206 JL |
3814 | if (retry_estale(error, lookup_flags)) { |
3815 | lookup_flags |= LOOKUP_REVAL; | |
3816 | goto retry; | |
3817 | } | |
1da177e4 LT |
3818 | return error; |
3819 | } | |
3820 | ||
3cdad428 | 3821 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3822 | { |
3823 | return do_rmdir(AT_FDCWD, pathname); | |
3824 | } | |
3825 | ||
b21996e3 BF |
3826 | /** |
3827 | * vfs_unlink - unlink a filesystem object | |
3828 | * @dir: parent directory | |
3829 | * @dentry: victim | |
3830 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3831 | * | |
3832 | * The caller must hold dir->i_mutex. | |
3833 | * | |
3834 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3835 | * return a reference to the inode in delegated_inode. The caller | |
3836 | * should then break the delegation on that inode and retry. Because | |
3837 | * breaking a delegation may take a long time, the caller should drop | |
3838 | * dir->i_mutex before doing so. | |
3839 | * | |
3840 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3841 | * be appropriate for callers that expect the underlying filesystem not | |
3842 | * to be NFS exported. | |
3843 | */ | |
3844 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3845 | { |
9accbb97 | 3846 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3847 | int error = may_delete(dir, dentry, 0); |
3848 | ||
3849 | if (error) | |
3850 | return error; | |
3851 | ||
acfa4380 | 3852 | if (!dir->i_op->unlink) |
1da177e4 LT |
3853 | return -EPERM; |
3854 | ||
5955102c | 3855 | inode_lock(target); |
8ed936b5 | 3856 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3857 | error = -EBUSY; |
3858 | else { | |
3859 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3860 | if (!error) { |
5a14696c BF |
3861 | error = try_break_deleg(target, delegated_inode); |
3862 | if (error) | |
b21996e3 | 3863 | goto out; |
1da177e4 | 3864 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3865 | if (!error) { |
d83c49f3 | 3866 | dont_mount(dentry); |
8ed936b5 EB |
3867 | detach_mounts(dentry); |
3868 | } | |
bec1052e | 3869 | } |
1da177e4 | 3870 | } |
b21996e3 | 3871 | out: |
5955102c | 3872 | inode_unlock(target); |
1da177e4 LT |
3873 | |
3874 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3875 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 3876 | fsnotify_link_count(target); |
e234f35c | 3877 | d_delete(dentry); |
1da177e4 | 3878 | } |
0eeca283 | 3879 | |
1da177e4 LT |
3880 | return error; |
3881 | } | |
4d359507 | 3882 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
3883 | |
3884 | /* | |
3885 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3886 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3887 | * writeout happening, and we don't want to prevent access to the directory |
3888 | * while waiting on the I/O. | |
3889 | */ | |
5590ff0d | 3890 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3891 | { |
2ad94ae6 | 3892 | int error; |
91a27b2a | 3893 | struct filename *name; |
1da177e4 | 3894 | struct dentry *dentry; |
f5beed75 AV |
3895 | struct path path; |
3896 | struct qstr last; | |
3897 | int type; | |
1da177e4 | 3898 | struct inode *inode = NULL; |
b21996e3 | 3899 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
3900 | unsigned int lookup_flags = 0; |
3901 | retry: | |
f5beed75 AV |
3902 | name = user_path_parent(dfd, pathname, |
3903 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3904 | if (IS_ERR(name)) |
3905 | return PTR_ERR(name); | |
2ad94ae6 | 3906 | |
1da177e4 | 3907 | error = -EISDIR; |
f5beed75 | 3908 | if (type != LAST_NORM) |
1da177e4 | 3909 | goto exit1; |
0612d9fb | 3910 | |
f5beed75 | 3911 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3912 | if (error) |
3913 | goto exit1; | |
b21996e3 | 3914 | retry_deleg: |
5955102c | 3915 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3916 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 LT |
3917 | error = PTR_ERR(dentry); |
3918 | if (!IS_ERR(dentry)) { | |
3919 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 3920 | if (last.name[last.len]) |
50338b88 | 3921 | goto slashes; |
1da177e4 | 3922 | inode = dentry->d_inode; |
b18825a7 | 3923 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
3924 | goto slashes; |
3925 | ihold(inode); | |
f5beed75 | 3926 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 3927 | if (error) |
c30dabfe | 3928 | goto exit2; |
f5beed75 | 3929 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 3930 | exit2: |
1da177e4 LT |
3931 | dput(dentry); |
3932 | } | |
5955102c | 3933 | inode_unlock(path.dentry->d_inode); |
1da177e4 LT |
3934 | if (inode) |
3935 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
3936 | inode = NULL; |
3937 | if (delegated_inode) { | |
5a14696c | 3938 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
3939 | if (!error) |
3940 | goto retry_deleg; | |
3941 | } | |
f5beed75 | 3942 | mnt_drop_write(path.mnt); |
1da177e4 | 3943 | exit1: |
f5beed75 | 3944 | path_put(&path); |
1da177e4 | 3945 | putname(name); |
5d18f813 JL |
3946 | if (retry_estale(error, lookup_flags)) { |
3947 | lookup_flags |= LOOKUP_REVAL; | |
3948 | inode = NULL; | |
3949 | goto retry; | |
3950 | } | |
1da177e4 LT |
3951 | return error; |
3952 | ||
3953 | slashes: | |
b18825a7 DH |
3954 | if (d_is_negative(dentry)) |
3955 | error = -ENOENT; | |
44b1d530 | 3956 | else if (d_is_dir(dentry)) |
b18825a7 DH |
3957 | error = -EISDIR; |
3958 | else | |
3959 | error = -ENOTDIR; | |
1da177e4 LT |
3960 | goto exit2; |
3961 | } | |
3962 | ||
2e4d0924 | 3963 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3964 | { |
3965 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3966 | return -EINVAL; | |
3967 | ||
3968 | if (flag & AT_REMOVEDIR) | |
3969 | return do_rmdir(dfd, pathname); | |
3970 | ||
3971 | return do_unlinkat(dfd, pathname); | |
3972 | } | |
3973 | ||
3480b257 | 3974 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3975 | { |
3976 | return do_unlinkat(AT_FDCWD, pathname); | |
3977 | } | |
3978 | ||
db2e747b | 3979 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3980 | { |
a95164d9 | 3981 | int error = may_create(dir, dentry); |
1da177e4 LT |
3982 | |
3983 | if (error) | |
3984 | return error; | |
3985 | ||
acfa4380 | 3986 | if (!dir->i_op->symlink) |
1da177e4 LT |
3987 | return -EPERM; |
3988 | ||
3989 | error = security_inode_symlink(dir, dentry, oldname); | |
3990 | if (error) | |
3991 | return error; | |
3992 | ||
1da177e4 | 3993 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3994 | if (!error) |
f38aa942 | 3995 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3996 | return error; |
3997 | } | |
4d359507 | 3998 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 3999 | |
2e4d0924 HC |
4000 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
4001 | int, newdfd, const char __user *, newname) | |
1da177e4 | 4002 | { |
2ad94ae6 | 4003 | int error; |
91a27b2a | 4004 | struct filename *from; |
6902d925 | 4005 | struct dentry *dentry; |
dae6ad8f | 4006 | struct path path; |
f46d3567 | 4007 | unsigned int lookup_flags = 0; |
1da177e4 LT |
4008 | |
4009 | from = getname(oldname); | |
2ad94ae6 | 4010 | if (IS_ERR(from)) |
1da177e4 | 4011 | return PTR_ERR(from); |
f46d3567 JL |
4012 | retry: |
4013 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
4014 | error = PTR_ERR(dentry); |
4015 | if (IS_ERR(dentry)) | |
dae6ad8f | 4016 | goto out_putname; |
6902d925 | 4017 | |
91a27b2a | 4018 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 4019 | if (!error) |
91a27b2a | 4020 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 4021 | done_path_create(&path, dentry); |
f46d3567 JL |
4022 | if (retry_estale(error, lookup_flags)) { |
4023 | lookup_flags |= LOOKUP_REVAL; | |
4024 | goto retry; | |
4025 | } | |
6902d925 | 4026 | out_putname: |
1da177e4 LT |
4027 | putname(from); |
4028 | return error; | |
4029 | } | |
4030 | ||
3480b257 | 4031 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
4032 | { |
4033 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
4034 | } | |
4035 | ||
146a8595 BF |
4036 | /** |
4037 | * vfs_link - create a new link | |
4038 | * @old_dentry: object to be linked | |
4039 | * @dir: new parent | |
4040 | * @new_dentry: where to create the new link | |
4041 | * @delegated_inode: returns inode needing a delegation break | |
4042 | * | |
4043 | * The caller must hold dir->i_mutex | |
4044 | * | |
4045 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
4046 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
4047 | * inode in delegated_inode. The caller should then break the delegation | |
4048 | * and retry. Because breaking a delegation may take a long time, the | |
4049 | * caller should drop the i_mutex before doing so. | |
4050 | * | |
4051 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4052 | * be appropriate for callers that expect the underlying filesystem not | |
4053 | * to be NFS exported. | |
4054 | */ | |
4055 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
4056 | { |
4057 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 4058 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
4059 | int error; |
4060 | ||
4061 | if (!inode) | |
4062 | return -ENOENT; | |
4063 | ||
a95164d9 | 4064 | error = may_create(dir, new_dentry); |
1da177e4 LT |
4065 | if (error) |
4066 | return error; | |
4067 | ||
4068 | if (dir->i_sb != inode->i_sb) | |
4069 | return -EXDEV; | |
4070 | ||
4071 | /* | |
4072 | * A link to an append-only or immutable file cannot be created. | |
4073 | */ | |
4074 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
4075 | return -EPERM; | |
acfa4380 | 4076 | if (!dir->i_op->link) |
1da177e4 | 4077 | return -EPERM; |
7e79eedb | 4078 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
4079 | return -EPERM; |
4080 | ||
4081 | error = security_inode_link(old_dentry, dir, new_dentry); | |
4082 | if (error) | |
4083 | return error; | |
4084 | ||
5955102c | 4085 | inode_lock(inode); |
aae8a97d | 4086 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 4087 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 4088 | error = -ENOENT; |
8de52778 AV |
4089 | else if (max_links && inode->i_nlink >= max_links) |
4090 | error = -EMLINK; | |
146a8595 BF |
4091 | else { |
4092 | error = try_break_deleg(inode, delegated_inode); | |
4093 | if (!error) | |
4094 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
4095 | } | |
f4e0c30c AV |
4096 | |
4097 | if (!error && (inode->i_state & I_LINKABLE)) { | |
4098 | spin_lock(&inode->i_lock); | |
4099 | inode->i_state &= ~I_LINKABLE; | |
4100 | spin_unlock(&inode->i_lock); | |
4101 | } | |
5955102c | 4102 | inode_unlock(inode); |
e31e14ec | 4103 | if (!error) |
7e79eedb | 4104 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
4105 | return error; |
4106 | } | |
4d359507 | 4107 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
4108 | |
4109 | /* | |
4110 | * Hardlinks are often used in delicate situations. We avoid | |
4111 | * security-related surprises by not following symlinks on the | |
4112 | * newname. --KAB | |
4113 | * | |
4114 | * We don't follow them on the oldname either to be compatible | |
4115 | * with linux 2.0, and to avoid hard-linking to directories | |
4116 | * and other special files. --ADM | |
4117 | */ | |
2e4d0924 HC |
4118 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4119 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
4120 | { |
4121 | struct dentry *new_dentry; | |
dae6ad8f | 4122 | struct path old_path, new_path; |
146a8595 | 4123 | struct inode *delegated_inode = NULL; |
11a7b371 | 4124 | int how = 0; |
1da177e4 | 4125 | int error; |
1da177e4 | 4126 | |
11a7b371 | 4127 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4128 | return -EINVAL; |
11a7b371 | 4129 | /* |
f0cc6ffb LT |
4130 | * To use null names we require CAP_DAC_READ_SEARCH |
4131 | * This ensures that not everyone will be able to create | |
4132 | * handlink using the passed filedescriptor. | |
11a7b371 | 4133 | */ |
f0cc6ffb LT |
4134 | if (flags & AT_EMPTY_PATH) { |
4135 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4136 | return -ENOENT; | |
11a7b371 | 4137 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4138 | } |
11a7b371 AK |
4139 | |
4140 | if (flags & AT_SYMLINK_FOLLOW) | |
4141 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4142 | retry: |
11a7b371 | 4143 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4144 | if (error) |
2ad94ae6 AV |
4145 | return error; |
4146 | ||
442e31ca JL |
4147 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4148 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4149 | error = PTR_ERR(new_dentry); |
6902d925 | 4150 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4151 | goto out; |
4152 | ||
4153 | error = -EXDEV; | |
4154 | if (old_path.mnt != new_path.mnt) | |
4155 | goto out_dput; | |
800179c9 KC |
4156 | error = may_linkat(&old_path); |
4157 | if (unlikely(error)) | |
4158 | goto out_dput; | |
dae6ad8f | 4159 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4160 | if (error) |
a8104a9f | 4161 | goto out_dput; |
146a8595 | 4162 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4163 | out_dput: |
921a1650 | 4164 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4165 | if (delegated_inode) { |
4166 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4167 | if (!error) { |
4168 | path_put(&old_path); | |
146a8595 | 4169 | goto retry; |
d22e6338 | 4170 | } |
146a8595 | 4171 | } |
442e31ca | 4172 | if (retry_estale(error, how)) { |
d22e6338 | 4173 | path_put(&old_path); |
442e31ca JL |
4174 | how |= LOOKUP_REVAL; |
4175 | goto retry; | |
4176 | } | |
1da177e4 | 4177 | out: |
2d8f3038 | 4178 | path_put(&old_path); |
1da177e4 LT |
4179 | |
4180 | return error; | |
4181 | } | |
4182 | ||
3480b257 | 4183 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4184 | { |
c04030e1 | 4185 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4186 | } |
4187 | ||
bc27027a MS |
4188 | /** |
4189 | * vfs_rename - rename a filesystem object | |
4190 | * @old_dir: parent of source | |
4191 | * @old_dentry: source | |
4192 | * @new_dir: parent of destination | |
4193 | * @new_dentry: destination | |
4194 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4195 | * @flags: rename flags |
bc27027a MS |
4196 | * |
4197 | * The caller must hold multiple mutexes--see lock_rename()). | |
4198 | * | |
4199 | * If vfs_rename discovers a delegation in need of breaking at either | |
4200 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4201 | * reference to the inode in delegated_inode. The caller should then | |
4202 | * break the delegation and retry. Because breaking a delegation may | |
4203 | * take a long time, the caller should drop all locks before doing | |
4204 | * so. | |
4205 | * | |
4206 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4207 | * be appropriate for callers that expect the underlying filesystem not | |
4208 | * to be NFS exported. | |
4209 | * | |
1da177e4 LT |
4210 | * The worst of all namespace operations - renaming directory. "Perverted" |
4211 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4212 | * Problems: | |
d03b29a2 | 4213 | * a) we can get into loop creation. |
1da177e4 LT |
4214 | * b) race potential - two innocent renames can create a loop together. |
4215 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4216 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4217 | * story. |
6cedba89 BF |
4218 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4219 | * and source (if it is not a directory). | |
1b1dcc1b | 4220 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4221 | * whether the target exists). Solution: try to be smart with locking |
4222 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4223 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4224 | * move will be locked. Thus we can rank directories by the tree |
4225 | * (ancestors first) and rank all non-directories after them. | |
4226 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4227 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4228 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4229 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4230 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4231 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4232 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4233 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4234 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4235 | * locking]. |
4236 | */ | |
bc27027a MS |
4237 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4238 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4239 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4240 | { |
bc27027a MS |
4241 | int error; |
4242 | bool is_dir = d_is_dir(old_dentry); | |
4243 | const unsigned char *old_name; | |
4244 | struct inode *source = old_dentry->d_inode; | |
9055cba7 | 4245 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4246 | bool new_is_dir = false; |
4247 | unsigned max_links = new_dir->i_sb->s_max_links; | |
bc27027a MS |
4248 | |
4249 | if (source == target) | |
4250 | return 0; | |
4251 | ||
4252 | error = may_delete(old_dir, old_dentry, is_dir); | |
4253 | if (error) | |
4254 | return error; | |
4255 | ||
da1ce067 | 4256 | if (!target) { |
bc27027a | 4257 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4258 | } else { |
4259 | new_is_dir = d_is_dir(new_dentry); | |
4260 | ||
4261 | if (!(flags & RENAME_EXCHANGE)) | |
4262 | error = may_delete(new_dir, new_dentry, is_dir); | |
4263 | else | |
4264 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4265 | } | |
bc27027a MS |
4266 | if (error) |
4267 | return error; | |
4268 | ||
7177a9c4 | 4269 | if (!old_dir->i_op->rename && !old_dir->i_op->rename2) |
bc27027a | 4270 | return -EPERM; |
1da177e4 | 4271 | |
520c8b16 MS |
4272 | if (flags && !old_dir->i_op->rename2) |
4273 | return -EINVAL; | |
4274 | ||
1da177e4 LT |
4275 | /* |
4276 | * If we are going to change the parent - check write permissions, | |
4277 | * we'll need to flip '..'. | |
4278 | */ | |
da1ce067 MS |
4279 | if (new_dir != old_dir) { |
4280 | if (is_dir) { | |
4281 | error = inode_permission(source, MAY_WRITE); | |
4282 | if (error) | |
4283 | return error; | |
4284 | } | |
4285 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4286 | error = inode_permission(target, MAY_WRITE); | |
4287 | if (error) | |
4288 | return error; | |
4289 | } | |
1da177e4 LT |
4290 | } |
4291 | ||
0b3974eb MS |
4292 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4293 | flags); | |
1da177e4 LT |
4294 | if (error) |
4295 | return error; | |
4296 | ||
bc27027a | 4297 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
1d2ef590 | 4298 | dget(new_dentry); |
da1ce067 | 4299 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4300 | lock_two_nondirectories(source, target); |
4301 | else if (target) | |
5955102c | 4302 | inode_lock(target); |
9055cba7 SW |
4303 | |
4304 | error = -EBUSY; | |
7af1364f | 4305 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4306 | goto out; |
4307 | ||
da1ce067 | 4308 | if (max_links && new_dir != old_dir) { |
bc27027a | 4309 | error = -EMLINK; |
da1ce067 | 4310 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4311 | goto out; |
da1ce067 MS |
4312 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4313 | old_dir->i_nlink >= max_links) | |
4314 | goto out; | |
4315 | } | |
4316 | if (is_dir && !(flags & RENAME_EXCHANGE) && target) | |
4317 | shrink_dcache_parent(new_dentry); | |
4318 | if (!is_dir) { | |
bc27027a | 4319 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4320 | if (error) |
4321 | goto out; | |
da1ce067 MS |
4322 | } |
4323 | if (target && !new_is_dir) { | |
4324 | error = try_break_deleg(target, delegated_inode); | |
4325 | if (error) | |
4326 | goto out; | |
8e6d782c | 4327 | } |
7177a9c4 | 4328 | if (!old_dir->i_op->rename2) { |
520c8b16 MS |
4329 | error = old_dir->i_op->rename(old_dir, old_dentry, |
4330 | new_dir, new_dentry); | |
4331 | } else { | |
7177a9c4 | 4332 | WARN_ON(old_dir->i_op->rename != NULL); |
520c8b16 MS |
4333 | error = old_dir->i_op->rename2(old_dir, old_dentry, |
4334 | new_dir, new_dentry, flags); | |
4335 | } | |
51892bbb SW |
4336 | if (error) |
4337 | goto out; | |
4338 | ||
da1ce067 | 4339 | if (!(flags & RENAME_EXCHANGE) && target) { |
bc27027a MS |
4340 | if (is_dir) |
4341 | target->i_flags |= S_DEAD; | |
51892bbb | 4342 | dont_mount(new_dentry); |
8ed936b5 | 4343 | detach_mounts(new_dentry); |
bc27027a | 4344 | } |
da1ce067 MS |
4345 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4346 | if (!(flags & RENAME_EXCHANGE)) | |
4347 | d_move(old_dentry, new_dentry); | |
4348 | else | |
4349 | d_exchange(old_dentry, new_dentry); | |
4350 | } | |
51892bbb | 4351 | out: |
da1ce067 | 4352 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4353 | unlock_two_nondirectories(source, target); |
4354 | else if (target) | |
5955102c | 4355 | inode_unlock(target); |
1da177e4 | 4356 | dput(new_dentry); |
da1ce067 | 4357 | if (!error) { |
123df294 | 4358 | fsnotify_move(old_dir, new_dir, old_name, is_dir, |
da1ce067 MS |
4359 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4360 | if (flags & RENAME_EXCHANGE) { | |
4361 | fsnotify_move(new_dir, old_dir, old_dentry->d_name.name, | |
4362 | new_is_dir, NULL, new_dentry); | |
4363 | } | |
4364 | } | |
0eeca283 RL |
4365 | fsnotify_oldname_free(old_name); |
4366 | ||
1da177e4 LT |
4367 | return error; |
4368 | } | |
4d359507 | 4369 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4370 | |
520c8b16 MS |
4371 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4372 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
1da177e4 | 4373 | { |
2ad94ae6 AV |
4374 | struct dentry *old_dentry, *new_dentry; |
4375 | struct dentry *trap; | |
f5beed75 AV |
4376 | struct path old_path, new_path; |
4377 | struct qstr old_last, new_last; | |
4378 | int old_type, new_type; | |
8e6d782c | 4379 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4380 | struct filename *from; |
4381 | struct filename *to; | |
f5beed75 | 4382 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4383 | bool should_retry = false; |
2ad94ae6 | 4384 | int error; |
520c8b16 | 4385 | |
0d7a8555 | 4386 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4387 | return -EINVAL; |
4388 | ||
0d7a8555 MS |
4389 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4390 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4391 | return -EINVAL; |
4392 | ||
0d7a8555 MS |
4393 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4394 | return -EPERM; | |
4395 | ||
f5beed75 AV |
4396 | if (flags & RENAME_EXCHANGE) |
4397 | target_flags = 0; | |
4398 | ||
c6a94284 | 4399 | retry: |
f5beed75 AV |
4400 | from = user_path_parent(olddfd, oldname, |
4401 | &old_path, &old_last, &old_type, lookup_flags); | |
91a27b2a JL |
4402 | if (IS_ERR(from)) { |
4403 | error = PTR_ERR(from); | |
1da177e4 | 4404 | goto exit; |
91a27b2a | 4405 | } |
1da177e4 | 4406 | |
f5beed75 AV |
4407 | to = user_path_parent(newdfd, newname, |
4408 | &new_path, &new_last, &new_type, lookup_flags); | |
91a27b2a JL |
4409 | if (IS_ERR(to)) { |
4410 | error = PTR_ERR(to); | |
1da177e4 | 4411 | goto exit1; |
91a27b2a | 4412 | } |
1da177e4 LT |
4413 | |
4414 | error = -EXDEV; | |
f5beed75 | 4415 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4416 | goto exit2; |
4417 | ||
1da177e4 | 4418 | error = -EBUSY; |
f5beed75 | 4419 | if (old_type != LAST_NORM) |
1da177e4 LT |
4420 | goto exit2; |
4421 | ||
0a7c3937 MS |
4422 | if (flags & RENAME_NOREPLACE) |
4423 | error = -EEXIST; | |
f5beed75 | 4424 | if (new_type != LAST_NORM) |
1da177e4 LT |
4425 | goto exit2; |
4426 | ||
f5beed75 | 4427 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4428 | if (error) |
4429 | goto exit2; | |
4430 | ||
8e6d782c | 4431 | retry_deleg: |
f5beed75 | 4432 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4433 | |
f5beed75 | 4434 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4435 | error = PTR_ERR(old_dentry); |
4436 | if (IS_ERR(old_dentry)) | |
4437 | goto exit3; | |
4438 | /* source must exist */ | |
4439 | error = -ENOENT; | |
b18825a7 | 4440 | if (d_is_negative(old_dentry)) |
1da177e4 | 4441 | goto exit4; |
f5beed75 | 4442 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4443 | error = PTR_ERR(new_dentry); |
4444 | if (IS_ERR(new_dentry)) | |
4445 | goto exit4; | |
4446 | error = -EEXIST; | |
4447 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4448 | goto exit5; | |
da1ce067 MS |
4449 | if (flags & RENAME_EXCHANGE) { |
4450 | error = -ENOENT; | |
4451 | if (d_is_negative(new_dentry)) | |
4452 | goto exit5; | |
4453 | ||
4454 | if (!d_is_dir(new_dentry)) { | |
4455 | error = -ENOTDIR; | |
f5beed75 | 4456 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4457 | goto exit5; |
4458 | } | |
4459 | } | |
1da177e4 | 4460 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4461 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4462 | error = -ENOTDIR; |
f5beed75 | 4463 | if (old_last.name[old_last.len]) |
0a7c3937 | 4464 | goto exit5; |
f5beed75 | 4465 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4466 | goto exit5; |
1da177e4 LT |
4467 | } |
4468 | /* source should not be ancestor of target */ | |
4469 | error = -EINVAL; | |
4470 | if (old_dentry == trap) | |
0a7c3937 | 4471 | goto exit5; |
1da177e4 | 4472 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4473 | if (!(flags & RENAME_EXCHANGE)) |
4474 | error = -ENOTEMPTY; | |
1da177e4 LT |
4475 | if (new_dentry == trap) |
4476 | goto exit5; | |
4477 | ||
f5beed75 AV |
4478 | error = security_path_rename(&old_path, old_dentry, |
4479 | &new_path, new_dentry, flags); | |
be6d3e56 | 4480 | if (error) |
c30dabfe | 4481 | goto exit5; |
f5beed75 AV |
4482 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4483 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4484 | &delegated_inode, flags); |
1da177e4 LT |
4485 | exit5: |
4486 | dput(new_dentry); | |
4487 | exit4: | |
4488 | dput(old_dentry); | |
4489 | exit3: | |
f5beed75 | 4490 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4491 | if (delegated_inode) { |
4492 | error = break_deleg_wait(&delegated_inode); | |
4493 | if (!error) | |
4494 | goto retry_deleg; | |
4495 | } | |
f5beed75 | 4496 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4497 | exit2: |
c6a94284 JL |
4498 | if (retry_estale(error, lookup_flags)) |
4499 | should_retry = true; | |
f5beed75 | 4500 | path_put(&new_path); |
2ad94ae6 | 4501 | putname(to); |
1da177e4 | 4502 | exit1: |
f5beed75 | 4503 | path_put(&old_path); |
1da177e4 | 4504 | putname(from); |
c6a94284 JL |
4505 | if (should_retry) { |
4506 | should_retry = false; | |
4507 | lookup_flags |= LOOKUP_REVAL; | |
4508 | goto retry; | |
4509 | } | |
2ad94ae6 | 4510 | exit: |
1da177e4 LT |
4511 | return error; |
4512 | } | |
4513 | ||
520c8b16 MS |
4514 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4515 | int, newdfd, const char __user *, newname) | |
4516 | { | |
4517 | return sys_renameat2(olddfd, oldname, newdfd, newname, 0); | |
4518 | } | |
4519 | ||
a26eab24 | 4520 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4521 | { |
520c8b16 | 4522 | return sys_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4523 | } |
4524 | ||
787fb6bc MS |
4525 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4526 | { | |
4527 | int error = may_create(dir, dentry); | |
4528 | if (error) | |
4529 | return error; | |
4530 | ||
4531 | if (!dir->i_op->mknod) | |
4532 | return -EPERM; | |
4533 | ||
4534 | return dir->i_op->mknod(dir, dentry, | |
4535 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4536 | } | |
4537 | EXPORT_SYMBOL(vfs_whiteout); | |
4538 | ||
5d826c84 | 4539 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4540 | { |
5d826c84 | 4541 | int len = PTR_ERR(link); |
1da177e4 LT |
4542 | if (IS_ERR(link)) |
4543 | goto out; | |
4544 | ||
4545 | len = strlen(link); | |
4546 | if (len > (unsigned) buflen) | |
4547 | len = buflen; | |
4548 | if (copy_to_user(buffer, link, len)) | |
4549 | len = -EFAULT; | |
4550 | out: | |
4551 | return len; | |
4552 | } | |
5d826c84 | 4553 | EXPORT_SYMBOL(readlink_copy); |
1da177e4 LT |
4554 | |
4555 | /* | |
4556 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
6b255391 AV |
4557 | * have ->get_link() not calling nd_jump_link(). Using (or not using) it |
4558 | * for any given inode is up to filesystem. | |
1da177e4 LT |
4559 | */ |
4560 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4561 | { | |
fceef393 | 4562 | DEFINE_DELAYED_CALL(done); |
5f2c4179 AV |
4563 | struct inode *inode = d_inode(dentry); |
4564 | const char *link = inode->i_link; | |
694a1764 | 4565 | int res; |
cc314eef | 4566 | |
d4dee48b | 4567 | if (!link) { |
fceef393 | 4568 | link = inode->i_op->get_link(dentry, inode, &done); |
d4dee48b AV |
4569 | if (IS_ERR(link)) |
4570 | return PTR_ERR(link); | |
4571 | } | |
680baacb | 4572 | res = readlink_copy(buffer, buflen, link); |
fceef393 | 4573 | do_delayed_call(&done); |
694a1764 | 4574 | return res; |
1da177e4 | 4575 | } |
4d359507 | 4576 | EXPORT_SYMBOL(generic_readlink); |
1da177e4 | 4577 | |
1da177e4 | 4578 | /* get the link contents into pagecache */ |
6b255391 | 4579 | const char *page_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 4580 | struct delayed_call *callback) |
1da177e4 | 4581 | { |
ebd09abb DG |
4582 | char *kaddr; |
4583 | struct page *page; | |
6b255391 AV |
4584 | struct address_space *mapping = inode->i_mapping; |
4585 | ||
d3883d4f AV |
4586 | if (!dentry) { |
4587 | page = find_get_page(mapping, 0); | |
4588 | if (!page) | |
4589 | return ERR_PTR(-ECHILD); | |
4590 | if (!PageUptodate(page)) { | |
4591 | put_page(page); | |
4592 | return ERR_PTR(-ECHILD); | |
4593 | } | |
4594 | } else { | |
4595 | page = read_mapping_page(mapping, 0, NULL); | |
4596 | if (IS_ERR(page)) | |
4597 | return (char*)page; | |
4598 | } | |
fceef393 | 4599 | set_delayed_call(callback, page_put_link, page); |
21fc61c7 AV |
4600 | BUG_ON(mapping_gfp_mask(mapping) & __GFP_HIGHMEM); |
4601 | kaddr = page_address(page); | |
6b255391 | 4602 | nd_terminate_link(kaddr, inode->i_size, PAGE_SIZE - 1); |
ebd09abb | 4603 | return kaddr; |
1da177e4 LT |
4604 | } |
4605 | ||
6b255391 | 4606 | EXPORT_SYMBOL(page_get_link); |
1da177e4 | 4607 | |
fceef393 | 4608 | void page_put_link(void *arg) |
1da177e4 | 4609 | { |
fceef393 | 4610 | put_page(arg); |
1da177e4 | 4611 | } |
4d359507 | 4612 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4613 | |
aa80deab AV |
4614 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) |
4615 | { | |
fceef393 | 4616 | DEFINE_DELAYED_CALL(done); |
6b255391 AV |
4617 | int res = readlink_copy(buffer, buflen, |
4618 | page_get_link(dentry, d_inode(dentry), | |
fceef393 AV |
4619 | &done)); |
4620 | do_delayed_call(&done); | |
aa80deab AV |
4621 | return res; |
4622 | } | |
4623 | EXPORT_SYMBOL(page_readlink); | |
4624 | ||
54566b2c NP |
4625 | /* |
4626 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4627 | */ | |
4628 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4629 | { |
4630 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4631 | struct page *page; |
afddba49 | 4632 | void *fsdata; |
beb497ab | 4633 | int err; |
54566b2c NP |
4634 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
4635 | if (nofs) | |
4636 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4637 | |
7e53cac4 | 4638 | retry: |
afddba49 | 4639 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4640 | flags, &page, &fsdata); |
1da177e4 | 4641 | if (err) |
afddba49 NP |
4642 | goto fail; |
4643 | ||
21fc61c7 | 4644 | memcpy(page_address(page), symname, len-1); |
afddba49 NP |
4645 | |
4646 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4647 | page, fsdata); | |
1da177e4 LT |
4648 | if (err < 0) |
4649 | goto fail; | |
afddba49 NP |
4650 | if (err < len-1) |
4651 | goto retry; | |
4652 | ||
1da177e4 LT |
4653 | mark_inode_dirty(inode); |
4654 | return 0; | |
1da177e4 LT |
4655 | fail: |
4656 | return err; | |
4657 | } | |
4d359507 | 4658 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4659 | |
0adb25d2 KK |
4660 | int page_symlink(struct inode *inode, const char *symname, int len) |
4661 | { | |
4662 | return __page_symlink(inode, symname, len, | |
c62d2555 | 4663 | !mapping_gfp_constraint(inode->i_mapping, __GFP_FS)); |
0adb25d2 | 4664 | } |
4d359507 | 4665 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4666 | |
92e1d5be | 4667 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 | 4668 | .readlink = generic_readlink, |
6b255391 | 4669 | .get_link = page_get_link, |
1da177e4 | 4670 | }; |
1da177e4 | 4671 | EXPORT_SYMBOL(page_symlink_inode_operations); |