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> | |
18 | #include <linux/module.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/fs.h> | |
21 | #include <linux/namei.h> | |
1da177e4 | 22 | #include <linux/pagemap.h> |
0eeca283 | 23 | #include <linux/fsnotify.h> |
1da177e4 LT |
24 | #include <linux/personality.h> |
25 | #include <linux/security.h> | |
6146f0d5 | 26 | #include <linux/ima.h> |
1da177e4 LT |
27 | #include <linux/syscalls.h> |
28 | #include <linux/mount.h> | |
29 | #include <linux/audit.h> | |
16f7e0fe | 30 | #include <linux/capability.h> |
834f2a4a | 31 | #include <linux/file.h> |
5590ff0d | 32 | #include <linux/fcntl.h> |
08ce5f16 | 33 | #include <linux/device_cgroup.h> |
5ad4e53b | 34 | #include <linux/fs_struct.h> |
1da177e4 LT |
35 | #include <asm/uaccess.h> |
36 | ||
e81e3f4d EP |
37 | #include "internal.h" |
38 | ||
1da177e4 LT |
39 | /* [Feb-1997 T. Schoebel-Theuer] |
40 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
41 | * were necessary because of omirr. The reason is that omirr needs | |
42 | * to know the _real_ pathname, not the user-supplied one, in case | |
43 | * of symlinks (and also when transname replacements occur). | |
44 | * | |
45 | * The new code replaces the old recursive symlink resolution with | |
46 | * an iterative one (in case of non-nested symlink chains). It does | |
47 | * this with calls to <fs>_follow_link(). | |
48 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
49 | * replaced with a single function lookup_dentry() that can handle all | |
50 | * the special cases of the former code. | |
51 | * | |
52 | * With the new dcache, the pathname is stored at each inode, at least as | |
53 | * long as the refcount of the inode is positive. As a side effect, the | |
54 | * size of the dcache depends on the inode cache and thus is dynamic. | |
55 | * | |
56 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
57 | * resolution to correspond with current state of the code. | |
58 | * | |
59 | * Note that the symlink resolution is not *completely* iterative. | |
60 | * There is still a significant amount of tail- and mid- recursion in | |
61 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
62 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
63 | * may return different results than <fs>_follow_link(). Many virtual | |
64 | * filesystems (including /proc) exhibit this behavior. | |
65 | */ | |
66 | ||
67 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
68 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
69 | * and the name already exists in form of a symlink, try to create the new | |
70 | * name indicated by the symlink. The old code always complained that the | |
71 | * name already exists, due to not following the symlink even if its target | |
72 | * is nonexistent. The new semantics affects also mknod() and link() when | |
73 | * the name is a symlink pointing to a non-existant name. | |
74 | * | |
75 | * I don't know which semantics is the right one, since I have no access | |
76 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
77 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
78 | * "old" one. Personally, I think the new semantics is much more logical. | |
79 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
80 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
81 | * and in the old Linux semantics. | |
82 | */ | |
83 | ||
84 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
85 | * semantics. See the comments in "open_namei" and "do_link" below. | |
86 | * | |
87 | * [10-Sep-98 Alan Modra] Another symlink change. | |
88 | */ | |
89 | ||
90 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
91 | * inside the path - always follow. | |
92 | * in the last component in creation/removal/renaming - never follow. | |
93 | * if LOOKUP_FOLLOW passed - follow. | |
94 | * if the pathname has trailing slashes - follow. | |
95 | * otherwise - don't follow. | |
96 | * (applied in that order). | |
97 | * | |
98 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
99 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
100 | * During the 2.4 we need to fix the userland stuff depending on it - | |
101 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
102 | * XEmacs seems to be relying on it... | |
103 | */ | |
104 | /* | |
105 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 106 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
107 | * any extra contention... |
108 | */ | |
109 | ||
110 | /* In order to reduce some races, while at the same time doing additional | |
111 | * checking and hopefully speeding things up, we copy filenames to the | |
112 | * kernel data space before using them.. | |
113 | * | |
114 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
115 | * PATH_MAX includes the nul terminator --RR. | |
116 | */ | |
858119e1 | 117 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
118 | { |
119 | int retval; | |
120 | unsigned long len = PATH_MAX; | |
121 | ||
122 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
123 | if ((unsigned long) filename >= TASK_SIZE) | |
124 | return -EFAULT; | |
125 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
126 | len = TASK_SIZE - (unsigned long) filename; | |
127 | } | |
128 | ||
129 | retval = strncpy_from_user(page, filename, len); | |
130 | if (retval > 0) { | |
131 | if (retval < len) | |
132 | return 0; | |
133 | return -ENAMETOOLONG; | |
134 | } else if (!retval) | |
135 | retval = -ENOENT; | |
136 | return retval; | |
137 | } | |
138 | ||
139 | char * getname(const char __user * filename) | |
140 | { | |
141 | char *tmp, *result; | |
142 | ||
143 | result = ERR_PTR(-ENOMEM); | |
144 | tmp = __getname(); | |
145 | if (tmp) { | |
146 | int retval = do_getname(filename, tmp); | |
147 | ||
148 | result = tmp; | |
149 | if (retval < 0) { | |
150 | __putname(tmp); | |
151 | result = ERR_PTR(retval); | |
152 | } | |
153 | } | |
154 | audit_getname(result); | |
155 | return result; | |
156 | } | |
157 | ||
158 | #ifdef CONFIG_AUDITSYSCALL | |
159 | void putname(const char *name) | |
160 | { | |
5ac3a9c2 | 161 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
162 | audit_putname(name); |
163 | else | |
164 | __putname(name); | |
165 | } | |
166 | EXPORT_SYMBOL(putname); | |
167 | #endif | |
168 | ||
5909ccaa LT |
169 | /* |
170 | * This does basic POSIX ACL permission checking | |
1da177e4 | 171 | */ |
5909ccaa | 172 | static int acl_permission_check(struct inode *inode, int mask, |
1da177e4 LT |
173 | int (*check_acl)(struct inode *inode, int mask)) |
174 | { | |
175 | umode_t mode = inode->i_mode; | |
176 | ||
e6305c43 AV |
177 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
178 | ||
da9592ed | 179 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
180 | mode >>= 6; |
181 | else { | |
182 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
183 | int error = check_acl(inode, mask); | |
5909ccaa | 184 | if (error != -EAGAIN) |
1da177e4 LT |
185 | return error; |
186 | } | |
187 | ||
188 | if (in_group_p(inode->i_gid)) | |
189 | mode >>= 3; | |
190 | } | |
191 | ||
192 | /* | |
193 | * If the DACs are ok we don't need any capability check. | |
194 | */ | |
e6305c43 | 195 | if ((mask & ~mode) == 0) |
1da177e4 | 196 | return 0; |
5909ccaa LT |
197 | return -EACCES; |
198 | } | |
199 | ||
200 | /** | |
201 | * generic_permission - check for access rights on a Posix-like filesystem | |
202 | * @inode: inode to check access rights for | |
203 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
204 | * @check_acl: optional callback to check for Posix ACLs | |
205 | * | |
206 | * Used to check for read/write/execute permissions on a file. | |
207 | * We use "fsuid" for this, letting us set arbitrary permissions | |
208 | * for filesystem access without changing the "normal" uids which | |
209 | * are used for other things.. | |
210 | */ | |
211 | int generic_permission(struct inode *inode, int mask, | |
212 | int (*check_acl)(struct inode *inode, int mask)) | |
213 | { | |
214 | int ret; | |
215 | ||
216 | /* | |
217 | * Do the basic POSIX ACL permission checks. | |
218 | */ | |
219 | ret = acl_permission_check(inode, mask, check_acl); | |
220 | if (ret != -EACCES) | |
221 | return ret; | |
1da177e4 | 222 | |
1da177e4 LT |
223 | /* |
224 | * Read/write DACs are always overridable. | |
225 | * Executable DACs are overridable if at least one exec bit is set. | |
226 | */ | |
f696a365 | 227 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
228 | if (capable(CAP_DAC_OVERRIDE)) |
229 | return 0; | |
230 | ||
231 | /* | |
232 | * Searching includes executable on directories, else just read. | |
233 | */ | |
7ea66001 | 234 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
1da177e4 LT |
235 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) |
236 | if (capable(CAP_DAC_READ_SEARCH)) | |
237 | return 0; | |
238 | ||
239 | return -EACCES; | |
240 | } | |
241 | ||
cb23beb5 CH |
242 | /** |
243 | * inode_permission - check for access rights to a given inode | |
244 | * @inode: inode to check permission on | |
245 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
246 | * | |
247 | * Used to check for read/write/execute permissions on an inode. | |
248 | * We use "fsuid" for this, letting us set arbitrary permissions | |
249 | * for filesystem access without changing the "normal" uids which | |
250 | * are used for other things. | |
251 | */ | |
f419a2e3 | 252 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 253 | { |
e6305c43 | 254 | int retval; |
1da177e4 LT |
255 | |
256 | if (mask & MAY_WRITE) { | |
22590e41 | 257 | umode_t mode = inode->i_mode; |
1da177e4 LT |
258 | |
259 | /* | |
260 | * Nobody gets write access to a read-only fs. | |
261 | */ | |
262 | if (IS_RDONLY(inode) && | |
263 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
264 | return -EROFS; | |
265 | ||
266 | /* | |
267 | * Nobody gets write access to an immutable file. | |
268 | */ | |
269 | if (IS_IMMUTABLE(inode)) | |
270 | return -EACCES; | |
271 | } | |
272 | ||
acfa4380 | 273 | if (inode->i_op->permission) |
b77b0646 | 274 | retval = inode->i_op->permission(inode, mask); |
f696a365 | 275 | else |
5909ccaa | 276 | retval = generic_permission(inode, mask, inode->i_op->check_acl); |
f696a365 | 277 | |
1da177e4 LT |
278 | if (retval) |
279 | return retval; | |
280 | ||
08ce5f16 SH |
281 | retval = devcgroup_inode_permission(inode, mask); |
282 | if (retval) | |
283 | return retval; | |
284 | ||
d09ca739 | 285 | return security_inode_permission(inode, mask); |
1da177e4 LT |
286 | } |
287 | ||
8c744fb8 CH |
288 | /** |
289 | * file_permission - check for additional access rights to a given file | |
290 | * @file: file to check access rights for | |
291 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
292 | * | |
293 | * Used to check for read/write/execute permissions on an already opened | |
294 | * file. | |
295 | * | |
296 | * Note: | |
297 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 298 | * be done using inode_permission(). |
8c744fb8 CH |
299 | */ |
300 | int file_permission(struct file *file, int mask) | |
301 | { | |
f419a2e3 | 302 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
303 | } |
304 | ||
1da177e4 LT |
305 | /* |
306 | * get_write_access() gets write permission for a file. | |
307 | * put_write_access() releases this write permission. | |
308 | * This is used for regular files. | |
309 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
310 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
311 | * can have the following values: | |
312 | * 0: no writers, no VM_DENYWRITE mappings | |
313 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
314 | * > 0: (i_writecount) users are writing to the file. | |
315 | * | |
316 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
317 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
318 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
319 | * to do the change if sign is wrong. Exclusion between them is provided by | |
320 | * the inode->i_lock spinlock. | |
321 | */ | |
322 | ||
323 | int get_write_access(struct inode * inode) | |
324 | { | |
325 | spin_lock(&inode->i_lock); | |
326 | if (atomic_read(&inode->i_writecount) < 0) { | |
327 | spin_unlock(&inode->i_lock); | |
328 | return -ETXTBSY; | |
329 | } | |
330 | atomic_inc(&inode->i_writecount); | |
331 | spin_unlock(&inode->i_lock); | |
332 | ||
333 | return 0; | |
334 | } | |
335 | ||
336 | int deny_write_access(struct file * file) | |
337 | { | |
0f7fc9e4 | 338 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
339 | |
340 | spin_lock(&inode->i_lock); | |
341 | if (atomic_read(&inode->i_writecount) > 0) { | |
342 | spin_unlock(&inode->i_lock); | |
343 | return -ETXTBSY; | |
344 | } | |
345 | atomic_dec(&inode->i_writecount); | |
346 | spin_unlock(&inode->i_lock); | |
347 | ||
348 | return 0; | |
349 | } | |
350 | ||
5dd784d0 JB |
351 | /** |
352 | * path_get - get a reference to a path | |
353 | * @path: path to get the reference to | |
354 | * | |
355 | * Given a path increment the reference count to the dentry and the vfsmount. | |
356 | */ | |
357 | void path_get(struct path *path) | |
358 | { | |
359 | mntget(path->mnt); | |
360 | dget(path->dentry); | |
361 | } | |
362 | EXPORT_SYMBOL(path_get); | |
363 | ||
1d957f9b JB |
364 | /** |
365 | * path_put - put a reference to a path | |
366 | * @path: path to put the reference to | |
367 | * | |
368 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
369 | */ | |
370 | void path_put(struct path *path) | |
1da177e4 | 371 | { |
1d957f9b JB |
372 | dput(path->dentry); |
373 | mntput(path->mnt); | |
1da177e4 | 374 | } |
1d957f9b | 375 | EXPORT_SYMBOL(path_put); |
1da177e4 | 376 | |
834f2a4a TM |
377 | /** |
378 | * release_open_intent - free up open intent resources | |
379 | * @nd: pointer to nameidata | |
380 | */ | |
381 | void release_open_intent(struct nameidata *nd) | |
382 | { | |
0f7fc9e4 | 383 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
384 | put_filp(nd->intent.open.file); |
385 | else | |
386 | fput(nd->intent.open.file); | |
387 | } | |
388 | ||
bcdc5e01 IK |
389 | static inline struct dentry * |
390 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
391 | { | |
392 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
393 | if (unlikely(status <= 0)) { | |
394 | /* | |
395 | * The dentry failed validation. | |
396 | * If d_revalidate returned 0 attempt to invalidate | |
397 | * the dentry otherwise d_revalidate is asking us | |
398 | * to return a fail status. | |
399 | */ | |
400 | if (!status) { | |
401 | if (!d_invalidate(dentry)) { | |
402 | dput(dentry); | |
403 | dentry = NULL; | |
404 | } | |
405 | } else { | |
406 | dput(dentry); | |
407 | dentry = ERR_PTR(status); | |
408 | } | |
409 | } | |
410 | return dentry; | |
411 | } | |
412 | ||
39159de2 JL |
413 | /* |
414 | * force_reval_path - force revalidation of a dentry | |
415 | * | |
416 | * In some situations the path walking code will trust dentries without | |
417 | * revalidating them. This causes problems for filesystems that depend on | |
418 | * d_revalidate to handle file opens (e.g. NFSv4). When FS_REVAL_DOT is set | |
419 | * (which indicates that it's possible for the dentry to go stale), force | |
420 | * a d_revalidate call before proceeding. | |
421 | * | |
422 | * Returns 0 if the revalidation was successful. If the revalidation fails, | |
423 | * either return the error returned by d_revalidate or -ESTALE if the | |
424 | * revalidation it just returned 0. If d_revalidate returns 0, we attempt to | |
425 | * invalidate the dentry. It's up to the caller to handle putting references | |
426 | * to the path if necessary. | |
427 | */ | |
428 | static int | |
429 | force_reval_path(struct path *path, struct nameidata *nd) | |
430 | { | |
431 | int status; | |
432 | struct dentry *dentry = path->dentry; | |
433 | ||
434 | /* | |
435 | * only check on filesystems where it's possible for the dentry to | |
436 | * become stale. It's assumed that if this flag is set then the | |
437 | * d_revalidate op will also be defined. | |
438 | */ | |
439 | if (!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) | |
440 | return 0; | |
441 | ||
442 | status = dentry->d_op->d_revalidate(dentry, nd); | |
443 | if (status > 0) | |
444 | return 0; | |
445 | ||
446 | if (!status) { | |
447 | d_invalidate(dentry); | |
448 | status = -ESTALE; | |
449 | } | |
450 | return status; | |
451 | } | |
452 | ||
1da177e4 | 453 | /* |
b75b5086 AV |
454 | * Short-cut version of permission(), for calling on directories |
455 | * during pathname resolution. Combines parts of permission() | |
456 | * and generic_permission(), and tests ONLY for MAY_EXEC permission. | |
1da177e4 LT |
457 | * |
458 | * If appropriate, check DAC only. If not appropriate, or | |
b75b5086 | 459 | * short-cut DAC fails, then call ->permission() to do more |
1da177e4 LT |
460 | * complete permission check. |
461 | */ | |
b75b5086 | 462 | static int exec_permission(struct inode *inode) |
1da177e4 | 463 | { |
5909ccaa | 464 | int ret; |
1da177e4 | 465 | |
cb9179ea | 466 | if (inode->i_op->permission) { |
5909ccaa | 467 | ret = inode->i_op->permission(inode, MAY_EXEC); |
cb9179ea LT |
468 | if (!ret) |
469 | goto ok; | |
470 | return ret; | |
471 | } | |
5909ccaa LT |
472 | ret = acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl); |
473 | if (!ret) | |
1da177e4 LT |
474 | goto ok; |
475 | ||
f1ac9f6b | 476 | if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
477 | goto ok; |
478 | ||
5909ccaa | 479 | return ret; |
1da177e4 | 480 | ok: |
b77b0646 | 481 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
482 | } |
483 | ||
2a737871 AV |
484 | static __always_inline void set_root(struct nameidata *nd) |
485 | { | |
f7ad3c6b MS |
486 | if (!nd->root.mnt) |
487 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
488 | } |
489 | ||
6de88d72 AV |
490 | static int link_path_walk(const char *, struct nameidata *); |
491 | ||
f1662356 | 492 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 493 | { |
1da177e4 LT |
494 | if (IS_ERR(link)) |
495 | goto fail; | |
496 | ||
497 | if (*link == '/') { | |
2a737871 | 498 | set_root(nd); |
1d957f9b | 499 | path_put(&nd->path); |
2a737871 AV |
500 | nd->path = nd->root; |
501 | path_get(&nd->root); | |
1da177e4 | 502 | } |
b4091d5f | 503 | |
def4af30 | 504 | return link_path_walk(link, nd); |
1da177e4 | 505 | fail: |
1d957f9b | 506 | path_put(&nd->path); |
1da177e4 LT |
507 | return PTR_ERR(link); |
508 | } | |
509 | ||
1d957f9b | 510 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
511 | { |
512 | dput(path->dentry); | |
4ac91378 | 513 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
514 | mntput(path->mnt); |
515 | } | |
516 | ||
517 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
518 | { | |
4ac91378 | 519 | dput(nd->path.dentry); |
9a229683 | 520 | if (nd->path.mnt != path->mnt) { |
4ac91378 | 521 | mntput(nd->path.mnt); |
9a229683 HS |
522 | nd->path.mnt = path->mnt; |
523 | } | |
4ac91378 | 524 | nd->path.dentry = path->dentry; |
051d3812 IK |
525 | } |
526 | ||
def4af30 AV |
527 | static __always_inline int |
528 | __do_follow_link(struct path *path, struct nameidata *nd, void **p) | |
1da177e4 LT |
529 | { |
530 | int error; | |
cd4e91d3 | 531 | struct dentry *dentry = path->dentry; |
1da177e4 | 532 | |
d671a1cb | 533 | touch_atime(path->mnt, dentry); |
1da177e4 | 534 | nd_set_link(nd, NULL); |
cd4e91d3 | 535 | |
4ac91378 | 536 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
537 | path_to_nameidata(path, nd); |
538 | dget(dentry); | |
539 | } | |
540 | mntget(path->mnt); | |
86acdca1 | 541 | nd->last_type = LAST_BIND; |
def4af30 AV |
542 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
543 | error = PTR_ERR(*p); | |
544 | if (!IS_ERR(*p)) { | |
1da177e4 | 545 | char *s = nd_get_link(nd); |
cc314eef | 546 | error = 0; |
1da177e4 LT |
547 | if (s) |
548 | error = __vfs_follow_link(nd, s); | |
39159de2 JL |
549 | else if (nd->last_type == LAST_BIND) { |
550 | error = force_reval_path(&nd->path, nd); | |
551 | if (error) | |
552 | path_put(&nd->path); | |
553 | } | |
1da177e4 | 554 | } |
1da177e4 LT |
555 | return error; |
556 | } | |
557 | ||
558 | /* | |
559 | * This limits recursive symlink follows to 8, while | |
560 | * limiting consecutive symlinks to 40. | |
561 | * | |
562 | * Without that kind of total limit, nasty chains of consecutive | |
563 | * symlinks can cause almost arbitrarily long lookups. | |
564 | */ | |
90ebe565 | 565 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 | 566 | { |
def4af30 | 567 | void *cookie; |
1da177e4 LT |
568 | int err = -ELOOP; |
569 | if (current->link_count >= MAX_NESTED_LINKS) | |
570 | goto loop; | |
571 | if (current->total_link_count >= 40) | |
572 | goto loop; | |
573 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
574 | cond_resched(); | |
90ebe565 | 575 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
576 | if (err) |
577 | goto loop; | |
578 | current->link_count++; | |
579 | current->total_link_count++; | |
580 | nd->depth++; | |
def4af30 AV |
581 | err = __do_follow_link(path, nd, &cookie); |
582 | if (!IS_ERR(cookie) && path->dentry->d_inode->i_op->put_link) | |
583 | path->dentry->d_inode->i_op->put_link(path->dentry, nd, cookie); | |
258fa999 | 584 | path_put(path); |
839d9f93 AV |
585 | current->link_count--; |
586 | nd->depth--; | |
1da177e4 LT |
587 | return err; |
588 | loop: | |
1d957f9b JB |
589 | path_put_conditional(path, nd); |
590 | path_put(&nd->path); | |
1da177e4 LT |
591 | return err; |
592 | } | |
593 | ||
bab77ebf | 594 | int follow_up(struct path *path) |
1da177e4 LT |
595 | { |
596 | struct vfsmount *parent; | |
597 | struct dentry *mountpoint; | |
99b7db7b NP |
598 | |
599 | br_read_lock(vfsmount_lock); | |
bab77ebf AV |
600 | parent = path->mnt->mnt_parent; |
601 | if (parent == path->mnt) { | |
99b7db7b | 602 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
603 | return 0; |
604 | } | |
605 | mntget(parent); | |
bab77ebf | 606 | mountpoint = dget(path->mnt->mnt_mountpoint); |
99b7db7b | 607 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
608 | dput(path->dentry); |
609 | path->dentry = mountpoint; | |
610 | mntput(path->mnt); | |
611 | path->mnt = parent; | |
1da177e4 LT |
612 | return 1; |
613 | } | |
614 | ||
b5c84bf6 NP |
615 | /* |
616 | * serialization is taken care of in namespace.c | |
1da177e4 | 617 | */ |
463ffb2e AV |
618 | static int __follow_mount(struct path *path) |
619 | { | |
620 | int res = 0; | |
621 | while (d_mountpoint(path->dentry)) { | |
1c755af4 | 622 | struct vfsmount *mounted = lookup_mnt(path); |
463ffb2e AV |
623 | if (!mounted) |
624 | break; | |
625 | dput(path->dentry); | |
626 | if (res) | |
627 | mntput(path->mnt); | |
628 | path->mnt = mounted; | |
629 | path->dentry = dget(mounted->mnt_root); | |
630 | res = 1; | |
631 | } | |
632 | return res; | |
633 | } | |
634 | ||
79ed0226 | 635 | static void follow_mount(struct path *path) |
1da177e4 | 636 | { |
79ed0226 | 637 | while (d_mountpoint(path->dentry)) { |
1c755af4 | 638 | struct vfsmount *mounted = lookup_mnt(path); |
1da177e4 LT |
639 | if (!mounted) |
640 | break; | |
79ed0226 AV |
641 | dput(path->dentry); |
642 | mntput(path->mnt); | |
643 | path->mnt = mounted; | |
644 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 | 645 | } |
1da177e4 LT |
646 | } |
647 | ||
9393bd07 | 648 | int follow_down(struct path *path) |
1da177e4 LT |
649 | { |
650 | struct vfsmount *mounted; | |
651 | ||
1c755af4 | 652 | mounted = lookup_mnt(path); |
1da177e4 | 653 | if (mounted) { |
9393bd07 AV |
654 | dput(path->dentry); |
655 | mntput(path->mnt); | |
656 | path->mnt = mounted; | |
657 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
658 | return 1; |
659 | } | |
660 | return 0; | |
661 | } | |
662 | ||
f1662356 | 663 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 664 | { |
2a737871 | 665 | set_root(nd); |
e518ddb7 | 666 | |
1da177e4 | 667 | while(1) { |
4ac91378 | 668 | struct dentry *old = nd->path.dentry; |
1da177e4 | 669 | |
2a737871 AV |
670 | if (nd->path.dentry == nd->root.dentry && |
671 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
672 | break; |
673 | } | |
4ac91378 | 674 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
675 | /* rare case of legitimate dget_parent()... */ |
676 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
677 | dput(old); |
678 | break; | |
679 | } | |
3088dd70 | 680 | if (!follow_up(&nd->path)) |
1da177e4 | 681 | break; |
1da177e4 | 682 | } |
79ed0226 | 683 | follow_mount(&nd->path); |
1da177e4 LT |
684 | } |
685 | ||
baa03890 NP |
686 | /* |
687 | * Allocate a dentry with name and parent, and perform a parent | |
688 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
689 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
690 | * have verified that no child exists while under i_mutex. | |
691 | */ | |
692 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
693 | struct qstr *name, struct nameidata *nd) | |
694 | { | |
695 | struct inode *inode = parent->d_inode; | |
696 | struct dentry *dentry; | |
697 | struct dentry *old; | |
698 | ||
699 | /* Don't create child dentry for a dead directory. */ | |
700 | if (unlikely(IS_DEADDIR(inode))) | |
701 | return ERR_PTR(-ENOENT); | |
702 | ||
703 | dentry = d_alloc(parent, name); | |
704 | if (unlikely(!dentry)) | |
705 | return ERR_PTR(-ENOMEM); | |
706 | ||
707 | old = inode->i_op->lookup(inode, dentry, nd); | |
708 | if (unlikely(old)) { | |
709 | dput(dentry); | |
710 | dentry = old; | |
711 | } | |
712 | return dentry; | |
713 | } | |
714 | ||
1da177e4 LT |
715 | /* |
716 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
717 | * small and for now I'd prefer to have fast path as straight as possible. | |
718 | * It _is_ time-critical. | |
719 | */ | |
720 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
721 | struct path *path) | |
722 | { | |
4ac91378 | 723 | struct vfsmount *mnt = nd->path.mnt; |
6e6b1bd1 AV |
724 | struct dentry *dentry, *parent; |
725 | struct inode *dir; | |
3cac260a AV |
726 | /* |
727 | * See if the low-level filesystem might want | |
728 | * to use its own hash.. | |
729 | */ | |
730 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { | |
b1e6a015 NP |
731 | int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, |
732 | nd->path.dentry->d_inode, name); | |
3cac260a AV |
733 | if (err < 0) |
734 | return err; | |
735 | } | |
1da177e4 | 736 | |
b04f784e NP |
737 | /* |
738 | * Rename seqlock is not required here because in the off chance | |
739 | * of a false negative due to a concurrent rename, we're going to | |
740 | * do the non-racy lookup, below. | |
741 | */ | |
3cac260a | 742 | dentry = __d_lookup(nd->path.dentry, name); |
1da177e4 LT |
743 | if (!dentry) |
744 | goto need_lookup; | |
2e2e88ea | 745 | found: |
1da177e4 LT |
746 | if (dentry->d_op && dentry->d_op->d_revalidate) |
747 | goto need_revalidate; | |
748 | done: | |
749 | path->mnt = mnt; | |
750 | path->dentry = dentry; | |
634ee701 | 751 | __follow_mount(path); |
1da177e4 LT |
752 | return 0; |
753 | ||
754 | need_lookup: | |
6e6b1bd1 AV |
755 | parent = nd->path.dentry; |
756 | dir = parent->d_inode; | |
757 | ||
758 | mutex_lock(&dir->i_mutex); | |
759 | /* | |
760 | * First re-do the cached lookup just in case it was created | |
b04f784e NP |
761 | * while we waited for the directory semaphore, or the first |
762 | * lookup failed due to an unrelated rename. | |
6e6b1bd1 | 763 | * |
b04f784e NP |
764 | * This could use version numbering or similar to avoid unnecessary |
765 | * cache lookups, but then we'd have to do the first lookup in the | |
766 | * non-racy way. However in the common case here, everything should | |
767 | * be hot in cache, so would it be a big win? | |
6e6b1bd1 AV |
768 | */ |
769 | dentry = d_lookup(parent, name); | |
baa03890 NP |
770 | if (likely(!dentry)) { |
771 | dentry = d_alloc_and_lookup(parent, name, nd); | |
6e6b1bd1 AV |
772 | mutex_unlock(&dir->i_mutex); |
773 | if (IS_ERR(dentry)) | |
774 | goto fail; | |
775 | goto done; | |
776 | } | |
6e6b1bd1 AV |
777 | /* |
778 | * Uhhuh! Nasty case: the cache was re-populated while | |
779 | * we waited on the semaphore. Need to revalidate. | |
780 | */ | |
781 | mutex_unlock(&dir->i_mutex); | |
2e2e88ea | 782 | goto found; |
1da177e4 LT |
783 | |
784 | need_revalidate: | |
bcdc5e01 IK |
785 | dentry = do_revalidate(dentry, nd); |
786 | if (!dentry) | |
787 | goto need_lookup; | |
788 | if (IS_ERR(dentry)) | |
789 | goto fail; | |
790 | goto done; | |
1da177e4 LT |
791 | |
792 | fail: | |
793 | return PTR_ERR(dentry); | |
794 | } | |
795 | ||
ac278a9c AV |
796 | /* |
797 | * This is a temporary kludge to deal with "automount" symlinks; proper | |
798 | * solution is to trigger them on follow_mount(), so that do_lookup() | |
799 | * would DTRT. To be killed before 2.6.34-final. | |
800 | */ | |
801 | static inline int follow_on_final(struct inode *inode, unsigned lookup_flags) | |
802 | { | |
803 | return inode && unlikely(inode->i_op->follow_link) && | |
804 | ((lookup_flags & LOOKUP_FOLLOW) || S_ISDIR(inode->i_mode)); | |
805 | } | |
806 | ||
1da177e4 LT |
807 | /* |
808 | * Name resolution. | |
ea3834d9 PM |
809 | * This is the basic name resolution function, turning a pathname into |
810 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 811 | * |
ea3834d9 PM |
812 | * Returns 0 and nd will have valid dentry and mnt on success. |
813 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 814 | */ |
6de88d72 | 815 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
816 | { |
817 | struct path next; | |
818 | struct inode *inode; | |
819 | int err; | |
820 | unsigned int lookup_flags = nd->flags; | |
821 | ||
822 | while (*name=='/') | |
823 | name++; | |
824 | if (!*name) | |
825 | goto return_reval; | |
826 | ||
4ac91378 | 827 | inode = nd->path.dentry->d_inode; |
1da177e4 | 828 | if (nd->depth) |
f55eab82 | 829 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
830 | |
831 | /* At this point we know we have a real path component. */ | |
832 | for(;;) { | |
833 | unsigned long hash; | |
834 | struct qstr this; | |
835 | unsigned int c; | |
836 | ||
cdce5d6b | 837 | nd->flags |= LOOKUP_CONTINUE; |
b75b5086 | 838 | err = exec_permission(inode); |
1da177e4 LT |
839 | if (err) |
840 | break; | |
841 | ||
842 | this.name = name; | |
843 | c = *(const unsigned char *)name; | |
844 | ||
845 | hash = init_name_hash(); | |
846 | do { | |
847 | name++; | |
848 | hash = partial_name_hash(c, hash); | |
849 | c = *(const unsigned char *)name; | |
850 | } while (c && (c != '/')); | |
851 | this.len = name - (const char *) this.name; | |
852 | this.hash = end_name_hash(hash); | |
853 | ||
854 | /* remove trailing slashes? */ | |
855 | if (!c) | |
856 | goto last_component; | |
857 | while (*++name == '/'); | |
858 | if (!*name) | |
859 | goto last_with_slashes; | |
860 | ||
861 | /* | |
862 | * "." and ".." are special - ".." especially so because it has | |
863 | * to be able to know about the current root directory and | |
864 | * parent relationships. | |
865 | */ | |
866 | if (this.name[0] == '.') switch (this.len) { | |
867 | default: | |
868 | break; | |
869 | case 2: | |
870 | if (this.name[1] != '.') | |
871 | break; | |
58c465eb | 872 | follow_dotdot(nd); |
4ac91378 | 873 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
874 | /* fallthrough */ |
875 | case 1: | |
876 | continue; | |
877 | } | |
1da177e4 LT |
878 | /* This does the actual lookups.. */ |
879 | err = do_lookup(nd, &this, &next); | |
880 | if (err) | |
881 | break; | |
1da177e4 LT |
882 | |
883 | err = -ENOENT; | |
884 | inode = next.dentry->d_inode; | |
885 | if (!inode) | |
886 | goto out_dput; | |
1da177e4 LT |
887 | |
888 | if (inode->i_op->follow_link) { | |
90ebe565 | 889 | err = do_follow_link(&next, nd); |
1da177e4 LT |
890 | if (err) |
891 | goto return_err; | |
892 | err = -ENOENT; | |
4ac91378 | 893 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
894 | if (!inode) |
895 | break; | |
09dd17d3 MS |
896 | } else |
897 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
898 | err = -ENOTDIR; |
899 | if (!inode->i_op->lookup) | |
900 | break; | |
901 | continue; | |
902 | /* here ends the main loop */ | |
903 | ||
904 | last_with_slashes: | |
905 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
906 | last_component: | |
f55eab82 TM |
907 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
908 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
909 | if (lookup_flags & LOOKUP_PARENT) |
910 | goto lookup_parent; | |
911 | if (this.name[0] == '.') switch (this.len) { | |
912 | default: | |
913 | break; | |
914 | case 2: | |
915 | if (this.name[1] != '.') | |
916 | break; | |
58c465eb | 917 | follow_dotdot(nd); |
4ac91378 | 918 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
919 | /* fallthrough */ |
920 | case 1: | |
921 | goto return_reval; | |
922 | } | |
1da177e4 LT |
923 | err = do_lookup(nd, &this, &next); |
924 | if (err) | |
925 | break; | |
1da177e4 | 926 | inode = next.dentry->d_inode; |
ac278a9c | 927 | if (follow_on_final(inode, lookup_flags)) { |
90ebe565 | 928 | err = do_follow_link(&next, nd); |
1da177e4 LT |
929 | if (err) |
930 | goto return_err; | |
4ac91378 | 931 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
932 | } else |
933 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
934 | err = -ENOENT; |
935 | if (!inode) | |
936 | break; | |
937 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
938 | err = -ENOTDIR; | |
acfa4380 | 939 | if (!inode->i_op->lookup) |
1da177e4 LT |
940 | break; |
941 | } | |
942 | goto return_base; | |
943 | lookup_parent: | |
944 | nd->last = this; | |
945 | nd->last_type = LAST_NORM; | |
946 | if (this.name[0] != '.') | |
947 | goto return_base; | |
948 | if (this.len == 1) | |
949 | nd->last_type = LAST_DOT; | |
950 | else if (this.len == 2 && this.name[1] == '.') | |
951 | nd->last_type = LAST_DOTDOT; | |
952 | else | |
953 | goto return_base; | |
954 | return_reval: | |
955 | /* | |
956 | * We bypassed the ordinary revalidation routines. | |
957 | * We may need to check the cached dentry for staleness. | |
958 | */ | |
4ac91378 JB |
959 | if (nd->path.dentry && nd->path.dentry->d_sb && |
960 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
961 | err = -ESTALE; |
962 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
963 | if (!nd->path.dentry->d_op->d_revalidate( |
964 | nd->path.dentry, nd)) | |
1da177e4 LT |
965 | break; |
966 | } | |
967 | return_base: | |
968 | return 0; | |
969 | out_dput: | |
1d957f9b | 970 | path_put_conditional(&next, nd); |
1da177e4 LT |
971 | break; |
972 | } | |
1d957f9b | 973 | path_put(&nd->path); |
1da177e4 LT |
974 | return_err: |
975 | return err; | |
976 | } | |
977 | ||
fc9b52cd | 978 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 979 | { |
6de88d72 AV |
980 | struct path save = nd->path; |
981 | int result; | |
982 | ||
1da177e4 | 983 | current->total_link_count = 0; |
6de88d72 AV |
984 | |
985 | /* make sure the stuff we saved doesn't go away */ | |
986 | path_get(&save); | |
987 | ||
988 | result = link_path_walk(name, nd); | |
989 | if (result == -ESTALE) { | |
990 | /* nd->path had been dropped */ | |
991 | current->total_link_count = 0; | |
992 | nd->path = save; | |
993 | path_get(&nd->path); | |
994 | nd->flags |= LOOKUP_REVAL; | |
995 | result = link_path_walk(name, nd); | |
996 | } | |
997 | ||
998 | path_put(&save); | |
999 | ||
1000 | return result; | |
1da177e4 LT |
1001 | } |
1002 | ||
9b4a9b14 | 1003 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1004 | { |
ea3834d9 | 1005 | int retval = 0; |
170aa3d0 UD |
1006 | int fput_needed; |
1007 | struct file *file; | |
1da177e4 LT |
1008 | |
1009 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1010 | nd->flags = flags; | |
1011 | nd->depth = 0; | |
2a737871 | 1012 | nd->root.mnt = NULL; |
1da177e4 | 1013 | |
1da177e4 | 1014 | if (*name=='/') { |
2a737871 AV |
1015 | set_root(nd); |
1016 | nd->path = nd->root; | |
1017 | path_get(&nd->root); | |
5590ff0d | 1018 | } else if (dfd == AT_FDCWD) { |
f7ad3c6b | 1019 | get_fs_pwd(current->fs, &nd->path); |
5590ff0d | 1020 | } else { |
5590ff0d UD |
1021 | struct dentry *dentry; |
1022 | ||
1023 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1024 | retval = -EBADF; |
1025 | if (!file) | |
6d09bb62 | 1026 | goto out_fail; |
5590ff0d | 1027 | |
0f7fc9e4 | 1028 | dentry = file->f_path.dentry; |
5590ff0d | 1029 | |
170aa3d0 UD |
1030 | retval = -ENOTDIR; |
1031 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1032 | goto fput_fail; |
5590ff0d UD |
1033 | |
1034 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1035 | if (retval) |
6d09bb62 | 1036 | goto fput_fail; |
5590ff0d | 1037 | |
5dd784d0 JB |
1038 | nd->path = file->f_path; |
1039 | path_get(&file->f_path); | |
5590ff0d UD |
1040 | |
1041 | fput_light(file, fput_needed); | |
1da177e4 | 1042 | } |
9b4a9b14 | 1043 | return 0; |
2dfdd266 | 1044 | |
9b4a9b14 AV |
1045 | fput_fail: |
1046 | fput_light(file, fput_needed); | |
1047 | out_fail: | |
1048 | return retval; | |
1049 | } | |
1050 | ||
1051 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1052 | static int do_path_lookup(int dfd, const char *name, | |
1053 | unsigned int flags, struct nameidata *nd) | |
1054 | { | |
1055 | int retval = path_init(dfd, name, flags, nd); | |
1056 | if (!retval) | |
1057 | retval = path_walk(name, nd); | |
4ac91378 JB |
1058 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1059 | nd->path.dentry->d_inode)) | |
1060 | audit_inode(name, nd->path.dentry); | |
2a737871 AV |
1061 | if (nd->root.mnt) { |
1062 | path_put(&nd->root); | |
1063 | nd->root.mnt = NULL; | |
1064 | } | |
170aa3d0 | 1065 | return retval; |
1da177e4 LT |
1066 | } |
1067 | ||
fc9b52cd | 1068 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1069 | struct nameidata *nd) |
1070 | { | |
1071 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1072 | } | |
1073 | ||
d1811465 AV |
1074 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1075 | { | |
1076 | struct nameidata nd; | |
1077 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1078 | if (!res) | |
1079 | *path = nd.path; | |
1080 | return res; | |
1081 | } | |
1082 | ||
16f18200 JJS |
1083 | /** |
1084 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1085 | * @dentry: pointer to dentry of the base directory | |
1086 | * @mnt: pointer to vfs mount of the base directory | |
1087 | * @name: pointer to file name | |
1088 | * @flags: lookup flags | |
1089 | * @nd: pointer to nameidata | |
1090 | */ | |
1091 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1092 | const char *name, unsigned int flags, | |
1093 | struct nameidata *nd) | |
1094 | { | |
1095 | int retval; | |
1096 | ||
1097 | /* same as do_path_lookup */ | |
1098 | nd->last_type = LAST_ROOT; | |
1099 | nd->flags = flags; | |
1100 | nd->depth = 0; | |
1101 | ||
c8e7f449 JB |
1102 | nd->path.dentry = dentry; |
1103 | nd->path.mnt = mnt; | |
1104 | path_get(&nd->path); | |
5b857119 AV |
1105 | nd->root = nd->path; |
1106 | path_get(&nd->root); | |
16f18200 JJS |
1107 | |
1108 | retval = path_walk(name, nd); | |
4ac91378 JB |
1109 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1110 | nd->path.dentry->d_inode)) | |
1111 | audit_inode(name, nd->path.dentry); | |
16f18200 | 1112 | |
5b857119 AV |
1113 | path_put(&nd->root); |
1114 | nd->root.mnt = NULL; | |
16f18200 | 1115 | |
2a737871 | 1116 | return retval; |
16f18200 JJS |
1117 | } |
1118 | ||
eead1911 CH |
1119 | static struct dentry *__lookup_hash(struct qstr *name, |
1120 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1121 | { |
81fca444 | 1122 | struct inode *inode = base->d_inode; |
057f6c01 | 1123 | struct dentry *dentry; |
1da177e4 LT |
1124 | int err; |
1125 | ||
81fca444 CH |
1126 | err = exec_permission(inode); |
1127 | if (err) | |
1128 | return ERR_PTR(err); | |
1da177e4 LT |
1129 | |
1130 | /* | |
1131 | * See if the low-level filesystem might want | |
1132 | * to use its own hash.. | |
1133 | */ | |
1134 | if (base->d_op && base->d_op->d_hash) { | |
b1e6a015 | 1135 | err = base->d_op->d_hash(base, inode, name); |
1da177e4 LT |
1136 | dentry = ERR_PTR(err); |
1137 | if (err < 0) | |
1138 | goto out; | |
1139 | } | |
1140 | ||
b04f784e NP |
1141 | /* |
1142 | * Don't bother with __d_lookup: callers are for creat as | |
1143 | * well as unlink, so a lot of the time it would cost | |
1144 | * a double lookup. | |
6e6b1bd1 | 1145 | */ |
b04f784e | 1146 | dentry = d_lookup(base, name); |
6e6b1bd1 AV |
1147 | |
1148 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) | |
1149 | dentry = do_revalidate(dentry, nd); | |
1150 | ||
baa03890 NP |
1151 | if (!dentry) |
1152 | dentry = d_alloc_and_lookup(base, name, nd); | |
1da177e4 LT |
1153 | out: |
1154 | return dentry; | |
1155 | } | |
1156 | ||
057f6c01 JM |
1157 | /* |
1158 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1159 | * needs parent already locked. Doesn't follow mounts. | |
1160 | * SMP-safe. | |
1161 | */ | |
eead1911 | 1162 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1163 | { |
4ac91378 | 1164 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1165 | } |
1166 | ||
eead1911 CH |
1167 | static int __lookup_one_len(const char *name, struct qstr *this, |
1168 | struct dentry *base, int len) | |
1da177e4 LT |
1169 | { |
1170 | unsigned long hash; | |
1da177e4 LT |
1171 | unsigned int c; |
1172 | ||
057f6c01 JM |
1173 | this->name = name; |
1174 | this->len = len; | |
1da177e4 | 1175 | if (!len) |
057f6c01 | 1176 | return -EACCES; |
1da177e4 LT |
1177 | |
1178 | hash = init_name_hash(); | |
1179 | while (len--) { | |
1180 | c = *(const unsigned char *)name++; | |
1181 | if (c == '/' || c == '\0') | |
057f6c01 | 1182 | return -EACCES; |
1da177e4 LT |
1183 | hash = partial_name_hash(c, hash); |
1184 | } | |
057f6c01 JM |
1185 | this->hash = end_name_hash(hash); |
1186 | return 0; | |
1187 | } | |
1da177e4 | 1188 | |
eead1911 | 1189 | /** |
a6b91919 | 1190 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1191 | * @name: pathname component to lookup |
1192 | * @base: base directory to lookup from | |
1193 | * @len: maximum length @len should be interpreted to | |
1194 | * | |
a6b91919 RD |
1195 | * Note that this routine is purely a helper for filesystem usage and should |
1196 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1197 | * nameidata argument is passed to the filesystem methods and a filesystem |
1198 | * using this helper needs to be prepared for that. | |
1199 | */ | |
057f6c01 JM |
1200 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1201 | { | |
1202 | int err; | |
1203 | struct qstr this; | |
1204 | ||
2f9092e1 DW |
1205 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1206 | ||
057f6c01 | 1207 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1208 | if (err) |
1209 | return ERR_PTR(err); | |
1210 | ||
49705b77 | 1211 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1212 | } |
1213 | ||
2d8f3038 AV |
1214 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1215 | struct path *path) | |
1da177e4 | 1216 | { |
2d8f3038 | 1217 | struct nameidata nd; |
1da177e4 LT |
1218 | char *tmp = getname(name); |
1219 | int err = PTR_ERR(tmp); | |
1da177e4 | 1220 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1221 | |
1222 | BUG_ON(flags & LOOKUP_PARENT); | |
1223 | ||
1224 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1225 | putname(tmp); |
2d8f3038 AV |
1226 | if (!err) |
1227 | *path = nd.path; | |
1da177e4 LT |
1228 | } |
1229 | return err; | |
1230 | } | |
1231 | ||
2ad94ae6 AV |
1232 | static int user_path_parent(int dfd, const char __user *path, |
1233 | struct nameidata *nd, char **name) | |
1234 | { | |
1235 | char *s = getname(path); | |
1236 | int error; | |
1237 | ||
1238 | if (IS_ERR(s)) | |
1239 | return PTR_ERR(s); | |
1240 | ||
1241 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1242 | if (error) | |
1243 | putname(s); | |
1244 | else | |
1245 | *name = s; | |
1246 | ||
1247 | return error; | |
1248 | } | |
1249 | ||
1da177e4 LT |
1250 | /* |
1251 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1252 | * minimal. | |
1253 | */ | |
1254 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1255 | { | |
da9592ed DH |
1256 | uid_t fsuid = current_fsuid(); |
1257 | ||
1da177e4 LT |
1258 | if (!(dir->i_mode & S_ISVTX)) |
1259 | return 0; | |
da9592ed | 1260 | if (inode->i_uid == fsuid) |
1da177e4 | 1261 | return 0; |
da9592ed | 1262 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1263 | return 0; |
1264 | return !capable(CAP_FOWNER); | |
1265 | } | |
1266 | ||
1267 | /* | |
1268 | * Check whether we can remove a link victim from directory dir, check | |
1269 | * whether the type of victim is right. | |
1270 | * 1. We can't do it if dir is read-only (done in permission()) | |
1271 | * 2. We should have write and exec permissions on dir | |
1272 | * 3. We can't remove anything from append-only dir | |
1273 | * 4. We can't do anything with immutable dir (done in permission()) | |
1274 | * 5. If the sticky bit on dir is set we should either | |
1275 | * a. be owner of dir, or | |
1276 | * b. be owner of victim, or | |
1277 | * c. have CAP_FOWNER capability | |
1278 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1279 | * links pointing to it. | |
1280 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1281 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1282 | * 9. We can't remove a root or mountpoint. | |
1283 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1284 | * nfs_async_unlink(). | |
1285 | */ | |
858119e1 | 1286 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1287 | { |
1288 | int error; | |
1289 | ||
1290 | if (!victim->d_inode) | |
1291 | return -ENOENT; | |
1292 | ||
1293 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1294 | audit_inode_child(victim, dir); |
1da177e4 | 1295 | |
f419a2e3 | 1296 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1297 | if (error) |
1298 | return error; | |
1299 | if (IS_APPEND(dir)) | |
1300 | return -EPERM; | |
1301 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1302 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1303 | return -EPERM; |
1304 | if (isdir) { | |
1305 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1306 | return -ENOTDIR; | |
1307 | if (IS_ROOT(victim)) | |
1308 | return -EBUSY; | |
1309 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1310 | return -EISDIR; | |
1311 | if (IS_DEADDIR(dir)) | |
1312 | return -ENOENT; | |
1313 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1314 | return -EBUSY; | |
1315 | return 0; | |
1316 | } | |
1317 | ||
1318 | /* Check whether we can create an object with dentry child in directory | |
1319 | * dir. | |
1320 | * 1. We can't do it if child already exists (open has special treatment for | |
1321 | * this case, but since we are inlined it's OK) | |
1322 | * 2. We can't do it if dir is read-only (done in permission()) | |
1323 | * 3. We should have write and exec permissions on dir | |
1324 | * 4. We can't do it if dir is immutable (done in permission()) | |
1325 | */ | |
a95164d9 | 1326 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1327 | { |
1328 | if (child->d_inode) | |
1329 | return -EEXIST; | |
1330 | if (IS_DEADDIR(dir)) | |
1331 | return -ENOENT; | |
f419a2e3 | 1332 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1333 | } |
1334 | ||
1da177e4 LT |
1335 | /* |
1336 | * p1 and p2 should be directories on the same fs. | |
1337 | */ | |
1338 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1339 | { | |
1340 | struct dentry *p; | |
1341 | ||
1342 | if (p1 == p2) { | |
f2eace23 | 1343 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1344 | return NULL; |
1345 | } | |
1346 | ||
a11f3a05 | 1347 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1348 | |
e2761a11 OH |
1349 | p = d_ancestor(p2, p1); |
1350 | if (p) { | |
1351 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1352 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1353 | return p; | |
1da177e4 LT |
1354 | } |
1355 | ||
e2761a11 OH |
1356 | p = d_ancestor(p1, p2); |
1357 | if (p) { | |
1358 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1359 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1360 | return p; | |
1da177e4 LT |
1361 | } |
1362 | ||
f2eace23 IM |
1363 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1364 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1365 | return NULL; |
1366 | } | |
1367 | ||
1368 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1369 | { | |
1b1dcc1b | 1370 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1371 | if (p1 != p2) { |
1b1dcc1b | 1372 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1373 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1374 | } |
1375 | } | |
1376 | ||
1377 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1378 | struct nameidata *nd) | |
1379 | { | |
a95164d9 | 1380 | int error = may_create(dir, dentry); |
1da177e4 LT |
1381 | |
1382 | if (error) | |
1383 | return error; | |
1384 | ||
acfa4380 | 1385 | if (!dir->i_op->create) |
1da177e4 LT |
1386 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1387 | mode &= S_IALLUGO; | |
1388 | mode |= S_IFREG; | |
1389 | error = security_inode_create(dir, dentry, mode); | |
1390 | if (error) | |
1391 | return error; | |
1da177e4 | 1392 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1393 | if (!error) |
f38aa942 | 1394 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1395 | return error; |
1396 | } | |
1397 | ||
3fb64190 | 1398 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1399 | { |
3fb64190 | 1400 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1401 | struct inode *inode = dentry->d_inode; |
1402 | int error; | |
1403 | ||
1404 | if (!inode) | |
1405 | return -ENOENT; | |
1406 | ||
c8fe8f30 CH |
1407 | switch (inode->i_mode & S_IFMT) { |
1408 | case S_IFLNK: | |
1da177e4 | 1409 | return -ELOOP; |
c8fe8f30 CH |
1410 | case S_IFDIR: |
1411 | if (acc_mode & MAY_WRITE) | |
1412 | return -EISDIR; | |
1413 | break; | |
1414 | case S_IFBLK: | |
1415 | case S_IFCHR: | |
3fb64190 | 1416 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1417 | return -EACCES; |
c8fe8f30 CH |
1418 | /*FALLTHRU*/ |
1419 | case S_IFIFO: | |
1420 | case S_IFSOCK: | |
1da177e4 | 1421 | flag &= ~O_TRUNC; |
c8fe8f30 | 1422 | break; |
4a3fd211 | 1423 | } |
b41572e9 | 1424 | |
3fb64190 | 1425 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1426 | if (error) |
1427 | return error; | |
6146f0d5 | 1428 | |
1da177e4 LT |
1429 | /* |
1430 | * An append-only file must be opened in append mode for writing. | |
1431 | */ | |
1432 | if (IS_APPEND(inode)) { | |
8737c930 | 1433 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 1434 | return -EPERM; |
1da177e4 | 1435 | if (flag & O_TRUNC) |
7715b521 | 1436 | return -EPERM; |
1da177e4 LT |
1437 | } |
1438 | ||
1439 | /* O_NOATIME can only be set by the owner or superuser */ | |
7715b521 AV |
1440 | if (flag & O_NOATIME && !is_owner_or_cap(inode)) |
1441 | return -EPERM; | |
1da177e4 LT |
1442 | |
1443 | /* | |
1444 | * Ensure there are no outstanding leases on the file. | |
1445 | */ | |
b65a9cfc | 1446 | return break_lease(inode, flag); |
7715b521 | 1447 | } |
1da177e4 | 1448 | |
7715b521 AV |
1449 | static int handle_truncate(struct path *path) |
1450 | { | |
1451 | struct inode *inode = path->dentry->d_inode; | |
1452 | int error = get_write_access(inode); | |
1453 | if (error) | |
1454 | return error; | |
1455 | /* | |
1456 | * Refuse to truncate files with mandatory locks held on them. | |
1457 | */ | |
1458 | error = locks_verify_locked(inode); | |
1459 | if (!error) | |
ea0d3ab2 | 1460 | error = security_path_truncate(path); |
7715b521 AV |
1461 | if (!error) { |
1462 | error = do_truncate(path->dentry, 0, | |
1463 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1464 | NULL); | |
1465 | } | |
1466 | put_write_access(inode); | |
acd0c935 | 1467 | return error; |
1da177e4 LT |
1468 | } |
1469 | ||
d57999e1 DH |
1470 | /* |
1471 | * Be careful about ever adding any more callers of this | |
1472 | * function. Its flags must be in the namei format, not | |
1473 | * what get passed to sys_open(). | |
1474 | */ | |
1475 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
8737c930 | 1476 | int open_flag, int mode) |
aab520e2 DH |
1477 | { |
1478 | int error; | |
4ac91378 | 1479 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1480 | |
1481 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1482 | mode &= ~current_umask(); |
be6d3e56 KT |
1483 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1484 | if (error) | |
1485 | goto out_unlock; | |
aab520e2 | 1486 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1487 | out_unlock: |
aab520e2 | 1488 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1489 | dput(nd->path.dentry); |
1490 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1491 | if (error) |
1492 | return error; | |
1493 | /* Don't check for write permission, don't truncate */ | |
8737c930 | 1494 | return may_open(&nd->path, 0, open_flag & ~O_TRUNC); |
aab520e2 DH |
1495 | } |
1496 | ||
d57999e1 DH |
1497 | /* |
1498 | * Note that while the flag value (low two bits) for sys_open means: | |
1499 | * 00 - read-only | |
1500 | * 01 - write-only | |
1501 | * 10 - read-write | |
1502 | * 11 - special | |
1503 | * it is changed into | |
1504 | * 00 - no permissions needed | |
1505 | * 01 - read-permission | |
1506 | * 10 - write-permission | |
1507 | * 11 - read-write | |
1508 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1509 | * This is more logical, and also allows the 00 "no perm needed" | |
1510 | * to be used for symlinks (where the permissions are checked | |
1511 | * later). | |
1512 | * | |
1513 | */ | |
1514 | static inline int open_to_namei_flags(int flag) | |
1515 | { | |
1516 | if ((flag+1) & O_ACCMODE) | |
1517 | flag++; | |
1518 | return flag; | |
1519 | } | |
1520 | ||
7715b521 | 1521 | static int open_will_truncate(int flag, struct inode *inode) |
4a3fd211 DH |
1522 | { |
1523 | /* | |
1524 | * We'll never write to the fs underlying | |
1525 | * a device file. | |
1526 | */ | |
1527 | if (special_file(inode->i_mode)) | |
1528 | return 0; | |
1529 | return (flag & O_TRUNC); | |
1530 | } | |
1531 | ||
648fa861 | 1532 | static struct file *finish_open(struct nameidata *nd, |
9a66179e | 1533 | int open_flag, int acc_mode) |
648fa861 AV |
1534 | { |
1535 | struct file *filp; | |
1536 | int will_truncate; | |
1537 | int error; | |
1538 | ||
9a66179e | 1539 | will_truncate = open_will_truncate(open_flag, nd->path.dentry->d_inode); |
648fa861 AV |
1540 | if (will_truncate) { |
1541 | error = mnt_want_write(nd->path.mnt); | |
1542 | if (error) | |
1543 | goto exit; | |
1544 | } | |
1545 | error = may_open(&nd->path, acc_mode, open_flag); | |
1546 | if (error) { | |
1547 | if (will_truncate) | |
1548 | mnt_drop_write(nd->path.mnt); | |
1549 | goto exit; | |
1550 | } | |
1551 | filp = nameidata_to_filp(nd); | |
1552 | if (!IS_ERR(filp)) { | |
1553 | error = ima_file_check(filp, acc_mode); | |
1554 | if (error) { | |
1555 | fput(filp); | |
1556 | filp = ERR_PTR(error); | |
1557 | } | |
1558 | } | |
1559 | if (!IS_ERR(filp)) { | |
648fa861 AV |
1560 | if (will_truncate) { |
1561 | error = handle_truncate(&nd->path); | |
1562 | if (error) { | |
1563 | fput(filp); | |
1564 | filp = ERR_PTR(error); | |
1565 | } | |
1566 | } | |
1567 | } | |
1568 | /* | |
1569 | * It is now safe to drop the mnt write | |
1570 | * because the filp has had a write taken | |
1571 | * on its behalf. | |
1572 | */ | |
1573 | if (will_truncate) | |
1574 | mnt_drop_write(nd->path.mnt); | |
d893f1bc | 1575 | path_put(&nd->path); |
648fa861 AV |
1576 | return filp; |
1577 | ||
1578 | exit: | |
1579 | if (!IS_ERR(nd->intent.open.file)) | |
1580 | release_open_intent(nd); | |
1581 | path_put(&nd->path); | |
1582 | return ERR_PTR(error); | |
1583 | } | |
1584 | ||
fb1cc555 | 1585 | static struct file *do_last(struct nameidata *nd, struct path *path, |
5b369df8 | 1586 | int open_flag, int acc_mode, |
3e297b61 | 1587 | int mode, const char *pathname) |
fb1cc555 | 1588 | { |
a1e28038 | 1589 | struct dentry *dir = nd->path.dentry; |
fb1cc555 | 1590 | struct file *filp; |
1f36f774 AV |
1591 | int error = -EISDIR; |
1592 | ||
1593 | switch (nd->last_type) { | |
1594 | case LAST_DOTDOT: | |
1595 | follow_dotdot(nd); | |
1596 | dir = nd->path.dentry; | |
176306f5 | 1597 | case LAST_DOT: |
1f36f774 AV |
1598 | if (nd->path.mnt->mnt_sb->s_type->fs_flags & FS_REVAL_DOT) { |
1599 | if (!dir->d_op->d_revalidate(dir, nd)) { | |
1600 | error = -ESTALE; | |
1601 | goto exit; | |
1602 | } | |
1603 | } | |
1604 | /* fallthrough */ | |
1f36f774 AV |
1605 | case LAST_ROOT: |
1606 | if (open_flag & O_CREAT) | |
1607 | goto exit; | |
1608 | /* fallthrough */ | |
1609 | case LAST_BIND: | |
1610 | audit_inode(pathname, dir); | |
67ee3ad2 | 1611 | goto ok; |
1f36f774 | 1612 | } |
67ee3ad2 | 1613 | |
1f36f774 AV |
1614 | /* trailing slashes? */ |
1615 | if (nd->last.name[nd->last.len]) { | |
1616 | if (open_flag & O_CREAT) | |
1617 | goto exit; | |
002baeec | 1618 | nd->flags |= LOOKUP_DIRECTORY | LOOKUP_FOLLOW; |
1f36f774 AV |
1619 | } |
1620 | ||
1621 | /* just plain open? */ | |
1622 | if (!(open_flag & O_CREAT)) { | |
1623 | error = do_lookup(nd, &nd->last, path); | |
1624 | if (error) | |
1625 | goto exit; | |
1626 | error = -ENOENT; | |
1627 | if (!path->dentry->d_inode) | |
1628 | goto exit_dput; | |
1629 | if (path->dentry->d_inode->i_op->follow_link) | |
1630 | return NULL; | |
1631 | error = -ENOTDIR; | |
3e297b61 AV |
1632 | if (nd->flags & LOOKUP_DIRECTORY) { |
1633 | if (!path->dentry->d_inode->i_op->lookup) | |
1634 | goto exit_dput; | |
1635 | } | |
1f36f774 AV |
1636 | path_to_nameidata(path, nd); |
1637 | audit_inode(pathname, nd->path.dentry); | |
1638 | goto ok; | |
1639 | } | |
a2c36b45 | 1640 | |
1f36f774 | 1641 | /* OK, it's O_CREAT */ |
a1e28038 AV |
1642 | mutex_lock(&dir->d_inode->i_mutex); |
1643 | ||
1644 | path->dentry = lookup_hash(nd); | |
1645 | path->mnt = nd->path.mnt; | |
1646 | ||
fb1cc555 AV |
1647 | error = PTR_ERR(path->dentry); |
1648 | if (IS_ERR(path->dentry)) { | |
1649 | mutex_unlock(&dir->d_inode->i_mutex); | |
1650 | goto exit; | |
1651 | } | |
1652 | ||
1653 | if (IS_ERR(nd->intent.open.file)) { | |
1654 | error = PTR_ERR(nd->intent.open.file); | |
1655 | goto exit_mutex_unlock; | |
1656 | } | |
1657 | ||
1658 | /* Negative dentry, just create the file */ | |
1659 | if (!path->dentry->d_inode) { | |
1660 | /* | |
1661 | * This write is needed to ensure that a | |
1662 | * ro->rw transition does not occur between | |
1663 | * the time when the file is created and when | |
1664 | * a permanent write count is taken through | |
1665 | * the 'struct file' in nameidata_to_filp(). | |
1666 | */ | |
1667 | error = mnt_want_write(nd->path.mnt); | |
1668 | if (error) | |
1669 | goto exit_mutex_unlock; | |
1670 | error = __open_namei_create(nd, path, open_flag, mode); | |
1671 | if (error) { | |
1672 | mnt_drop_write(nd->path.mnt); | |
1673 | goto exit; | |
1674 | } | |
1675 | filp = nameidata_to_filp(nd); | |
1676 | mnt_drop_write(nd->path.mnt); | |
d893f1bc | 1677 | path_put(&nd->path); |
fb1cc555 AV |
1678 | if (!IS_ERR(filp)) { |
1679 | error = ima_file_check(filp, acc_mode); | |
1680 | if (error) { | |
1681 | fput(filp); | |
1682 | filp = ERR_PTR(error); | |
1683 | } | |
1684 | } | |
1685 | return filp; | |
1686 | } | |
1687 | ||
1688 | /* | |
1689 | * It already exists. | |
1690 | */ | |
1691 | mutex_unlock(&dir->d_inode->i_mutex); | |
1692 | audit_inode(pathname, path->dentry); | |
1693 | ||
1694 | error = -EEXIST; | |
5b369df8 | 1695 | if (open_flag & O_EXCL) |
fb1cc555 AV |
1696 | goto exit_dput; |
1697 | ||
1698 | if (__follow_mount(path)) { | |
1699 | error = -ELOOP; | |
5b369df8 | 1700 | if (open_flag & O_NOFOLLOW) |
fb1cc555 AV |
1701 | goto exit_dput; |
1702 | } | |
1703 | ||
1704 | error = -ENOENT; | |
1705 | if (!path->dentry->d_inode) | |
1706 | goto exit_dput; | |
9e67f361 AV |
1707 | |
1708 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 1709 | return NULL; |
fb1cc555 AV |
1710 | |
1711 | path_to_nameidata(path, nd); | |
1712 | error = -EISDIR; | |
1713 | if (S_ISDIR(path->dentry->d_inode->i_mode)) | |
1714 | goto exit; | |
67ee3ad2 | 1715 | ok: |
9a66179e | 1716 | filp = finish_open(nd, open_flag, acc_mode); |
fb1cc555 AV |
1717 | return filp; |
1718 | ||
1719 | exit_mutex_unlock: | |
1720 | mutex_unlock(&dir->d_inode->i_mutex); | |
1721 | exit_dput: | |
1722 | path_put_conditional(path, nd); | |
1723 | exit: | |
1724 | if (!IS_ERR(nd->intent.open.file)) | |
1725 | release_open_intent(nd); | |
fb1cc555 AV |
1726 | path_put(&nd->path); |
1727 | return ERR_PTR(error); | |
1728 | } | |
1729 | ||
1da177e4 | 1730 | /* |
4a3fd211 DH |
1731 | * Note that the low bits of the passed in "open_flag" |
1732 | * are not the same as in the local variable "flag". See | |
1733 | * open_to_namei_flags() for more details. | |
1da177e4 | 1734 | */ |
a70e65df | 1735 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1736 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1737 | { |
4a3fd211 | 1738 | struct file *filp; |
a70e65df | 1739 | struct nameidata nd; |
6e8341a1 | 1740 | int error; |
9850c056 | 1741 | struct path path; |
1da177e4 | 1742 | int count = 0; |
d57999e1 | 1743 | int flag = open_to_namei_flags(open_flag); |
9850c056 | 1744 | int force_reval = 0; |
1f36f774 AV |
1745 | |
1746 | if (!(open_flag & O_CREAT)) | |
1747 | mode = 0; | |
1da177e4 | 1748 | |
b1085ba8 LS |
1749 | /* Must never be set by userspace */ |
1750 | open_flag &= ~FMODE_NONOTIFY; | |
1751 | ||
6b2f3d1f CH |
1752 | /* |
1753 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1754 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1755 | * always set instead of having to deal with possibly weird behaviour | |
1756 | * for malicious applications setting only __O_SYNC. | |
1757 | */ | |
1758 | if (open_flag & __O_SYNC) | |
1759 | open_flag |= O_DSYNC; | |
1760 | ||
6e8341a1 | 1761 | if (!acc_mode) |
6d125529 | 1762 | acc_mode = MAY_OPEN | ACC_MODE(open_flag); |
1da177e4 | 1763 | |
834f2a4a | 1764 | /* O_TRUNC implies we need access checks for write permissions */ |
4296e2cb | 1765 | if (open_flag & O_TRUNC) |
834f2a4a TM |
1766 | acc_mode |= MAY_WRITE; |
1767 | ||
1da177e4 LT |
1768 | /* Allow the LSM permission hook to distinguish append |
1769 | access from general write access. */ | |
4296e2cb | 1770 | if (open_flag & O_APPEND) |
1da177e4 LT |
1771 | acc_mode |= MAY_APPEND; |
1772 | ||
1f36f774 | 1773 | /* find the parent */ |
9850c056 | 1774 | reval: |
9b4a9b14 | 1775 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1776 | if (error) |
a70e65df | 1777 | return ERR_PTR(error); |
9850c056 AV |
1778 | if (force_reval) |
1779 | nd.flags |= LOOKUP_REVAL; | |
3866248e AV |
1780 | |
1781 | current->total_link_count = 0; | |
1782 | error = link_path_walk(pathname, &nd); | |
654f562c | 1783 | if (error) { |
10fa8e62 AV |
1784 | filp = ERR_PTR(error); |
1785 | goto out; | |
654f562c | 1786 | } |
1f36f774 | 1787 | if (unlikely(!audit_dummy_context()) && (open_flag & O_CREAT)) |
9b4a9b14 | 1788 | audit_inode(pathname, nd.path.dentry); |
1da177e4 LT |
1789 | |
1790 | /* | |
a2c36b45 | 1791 | * We have the parent and last component. |
1da177e4 | 1792 | */ |
1da177e4 | 1793 | |
8737f3a1 AV |
1794 | error = -ENFILE; |
1795 | filp = get_empty_filp(); | |
1796 | if (filp == NULL) | |
1797 | goto exit_parent; | |
1798 | nd.intent.open.file = filp; | |
482928d5 | 1799 | filp->f_flags = open_flag; |
8737f3a1 AV |
1800 | nd.intent.open.flags = flag; |
1801 | nd.intent.open.create_mode = mode; | |
a70e65df | 1802 | nd.flags &= ~LOOKUP_PARENT; |
1f36f774 AV |
1803 | nd.flags |= LOOKUP_OPEN; |
1804 | if (open_flag & O_CREAT) { | |
1805 | nd.flags |= LOOKUP_CREATE; | |
1806 | if (open_flag & O_EXCL) | |
1807 | nd.flags |= LOOKUP_EXCL; | |
1808 | } | |
3e297b61 AV |
1809 | if (open_flag & O_DIRECTORY) |
1810 | nd.flags |= LOOKUP_DIRECTORY; | |
002baeec JK |
1811 | if (!(open_flag & O_NOFOLLOW)) |
1812 | nd.flags |= LOOKUP_FOLLOW; | |
3e297b61 | 1813 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
806b681c | 1814 | while (unlikely(!filp)) { /* trailing symlink */ |
def4af30 | 1815 | struct path holder; |
1f36f774 | 1816 | struct inode *inode = path.dentry->d_inode; |
def4af30 | 1817 | void *cookie; |
806b681c | 1818 | error = -ELOOP; |
1f36f774 | 1819 | /* S_ISDIR part is a temporary automount kludge */ |
002baeec | 1820 | if (!(nd.flags & LOOKUP_FOLLOW) && !S_ISDIR(inode->i_mode)) |
1f36f774 AV |
1821 | goto exit_dput; |
1822 | if (count++ == 32) | |
806b681c AV |
1823 | goto exit_dput; |
1824 | /* | |
1825 | * This is subtle. Instead of calling do_follow_link() we do | |
1826 | * the thing by hands. The reason is that this way we have zero | |
1827 | * link_count and path_walk() (called from ->follow_link) | |
1828 | * honoring LOOKUP_PARENT. After that we have the parent and | |
1829 | * last component, i.e. we are in the same situation as after | |
1830 | * the first path_walk(). Well, almost - if the last component | |
1831 | * is normal we get its copy stored in nd->last.name and we will | |
1832 | * have to putname() it when we are done. Procfs-like symlinks | |
1833 | * just set LAST_BIND. | |
1834 | */ | |
1835 | nd.flags |= LOOKUP_PARENT; | |
1836 | error = security_inode_follow_link(path.dentry, &nd); | |
1837 | if (error) | |
1838 | goto exit_dput; | |
def4af30 AV |
1839 | error = __do_follow_link(&path, &nd, &cookie); |
1840 | if (unlikely(error)) { | |
806b681c | 1841 | /* nd.path had been dropped */ |
def4af30 AV |
1842 | if (!IS_ERR(cookie) && inode->i_op->put_link) |
1843 | inode->i_op->put_link(path.dentry, &nd, cookie); | |
1844 | path_put(&path); | |
806b681c AV |
1845 | release_open_intent(&nd); |
1846 | filp = ERR_PTR(error); | |
1847 | goto out; | |
1848 | } | |
def4af30 | 1849 | holder = path; |
806b681c | 1850 | nd.flags &= ~LOOKUP_PARENT; |
3e297b61 | 1851 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
def4af30 AV |
1852 | if (inode->i_op->put_link) |
1853 | inode->i_op->put_link(holder.dentry, &nd, cookie); | |
1854 | path_put(&holder); | |
806b681c | 1855 | } |
10fa8e62 | 1856 | out: |
2a737871 AV |
1857 | if (nd.root.mnt) |
1858 | path_put(&nd.root); | |
10fa8e62 AV |
1859 | if (filp == ERR_PTR(-ESTALE) && !force_reval) { |
1860 | force_reval = 1; | |
1861 | goto reval; | |
1862 | } | |
1863 | return filp; | |
1da177e4 | 1864 | |
806b681c AV |
1865 | exit_dput: |
1866 | path_put_conditional(&path, &nd); | |
1867 | if (!IS_ERR(nd.intent.open.file)) | |
a70e65df | 1868 | release_open_intent(&nd); |
806b681c AV |
1869 | exit_parent: |
1870 | path_put(&nd.path); | |
1871 | filp = ERR_PTR(error); | |
10fa8e62 | 1872 | goto out; |
1da177e4 LT |
1873 | } |
1874 | ||
a70e65df CH |
1875 | /** |
1876 | * filp_open - open file and return file pointer | |
1877 | * | |
1878 | * @filename: path to open | |
1879 | * @flags: open flags as per the open(2) second argument | |
1880 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1881 | * | |
1882 | * This is the helper to open a file from kernelspace if you really | |
1883 | * have to. But in generally you should not do this, so please move | |
1884 | * along, nothing to see here.. | |
1885 | */ | |
1886 | struct file *filp_open(const char *filename, int flags, int mode) | |
1887 | { | |
6e8341a1 | 1888 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1889 | } |
1890 | EXPORT_SYMBOL(filp_open); | |
1891 | ||
1da177e4 LT |
1892 | /** |
1893 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1894 | * @nd: nameidata info | |
1895 | * @is_dir: directory flag | |
1896 | * | |
1897 | * Simple function to lookup and return a dentry and create it | |
1898 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1899 | * |
4ac91378 | 1900 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1901 | */ |
1902 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1903 | { | |
c663e5d8 | 1904 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1905 | |
4ac91378 | 1906 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1907 | /* |
1908 | * Yucky last component or no last component at all? | |
1909 | * (foo/., foo/.., /////) | |
1910 | */ | |
1da177e4 LT |
1911 | if (nd->last_type != LAST_NORM) |
1912 | goto fail; | |
1913 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1914 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1915 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1916 | |
1917 | /* | |
1918 | * Do the final lookup. | |
1919 | */ | |
49705b77 | 1920 | dentry = lookup_hash(nd); |
1da177e4 LT |
1921 | if (IS_ERR(dentry)) |
1922 | goto fail; | |
c663e5d8 | 1923 | |
e9baf6e5 AV |
1924 | if (dentry->d_inode) |
1925 | goto eexist; | |
c663e5d8 CH |
1926 | /* |
1927 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1928 | * From the vfs_mknod() POV we just have a negative dentry - | |
1929 | * all is fine. Let's be bastards - you had / on the end, you've | |
1930 | * been asking for (non-existent) directory. -ENOENT for you. | |
1931 | */ | |
e9baf6e5 AV |
1932 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1933 | dput(dentry); | |
1934 | dentry = ERR_PTR(-ENOENT); | |
1935 | } | |
1da177e4 | 1936 | return dentry; |
e9baf6e5 | 1937 | eexist: |
1da177e4 | 1938 | dput(dentry); |
e9baf6e5 | 1939 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1940 | fail: |
1941 | return dentry; | |
1942 | } | |
f81a0bff | 1943 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1944 | |
1945 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1946 | { | |
a95164d9 | 1947 | int error = may_create(dir, dentry); |
1da177e4 LT |
1948 | |
1949 | if (error) | |
1950 | return error; | |
1951 | ||
1952 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1953 | return -EPERM; | |
1954 | ||
acfa4380 | 1955 | if (!dir->i_op->mknod) |
1da177e4 LT |
1956 | return -EPERM; |
1957 | ||
08ce5f16 SH |
1958 | error = devcgroup_inode_mknod(mode, dev); |
1959 | if (error) | |
1960 | return error; | |
1961 | ||
1da177e4 LT |
1962 | error = security_inode_mknod(dir, dentry, mode, dev); |
1963 | if (error) | |
1964 | return error; | |
1965 | ||
1da177e4 | 1966 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1967 | if (!error) |
f38aa942 | 1968 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1969 | return error; |
1970 | } | |
1971 | ||
463c3197 DH |
1972 | static int may_mknod(mode_t mode) |
1973 | { | |
1974 | switch (mode & S_IFMT) { | |
1975 | case S_IFREG: | |
1976 | case S_IFCHR: | |
1977 | case S_IFBLK: | |
1978 | case S_IFIFO: | |
1979 | case S_IFSOCK: | |
1980 | case 0: /* zero mode translates to S_IFREG */ | |
1981 | return 0; | |
1982 | case S_IFDIR: | |
1983 | return -EPERM; | |
1984 | default: | |
1985 | return -EINVAL; | |
1986 | } | |
1987 | } | |
1988 | ||
2e4d0924 HC |
1989 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
1990 | unsigned, dev) | |
1da177e4 | 1991 | { |
2ad94ae6 AV |
1992 | int error; |
1993 | char *tmp; | |
1994 | struct dentry *dentry; | |
1da177e4 LT |
1995 | struct nameidata nd; |
1996 | ||
1997 | if (S_ISDIR(mode)) | |
1998 | return -EPERM; | |
1da177e4 | 1999 | |
2ad94ae6 | 2000 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2001 | if (error) |
2ad94ae6 AV |
2002 | return error; |
2003 | ||
1da177e4 | 2004 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2005 | if (IS_ERR(dentry)) { |
2006 | error = PTR_ERR(dentry); | |
2007 | goto out_unlock; | |
2008 | } | |
4ac91378 | 2009 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2010 | mode &= ~current_umask(); |
463c3197 DH |
2011 | error = may_mknod(mode); |
2012 | if (error) | |
2013 | goto out_dput; | |
2014 | error = mnt_want_write(nd.path.mnt); | |
2015 | if (error) | |
2016 | goto out_dput; | |
be6d3e56 KT |
2017 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2018 | if (error) | |
2019 | goto out_drop_write; | |
463c3197 | 2020 | switch (mode & S_IFMT) { |
1da177e4 | 2021 | case 0: case S_IFREG: |
4ac91378 | 2022 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2023 | break; |
2024 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2025 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2026 | new_decode_dev(dev)); |
2027 | break; | |
2028 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2029 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2030 | break; |
1da177e4 | 2031 | } |
be6d3e56 | 2032 | out_drop_write: |
463c3197 DH |
2033 | mnt_drop_write(nd.path.mnt); |
2034 | out_dput: | |
2035 | dput(dentry); | |
2036 | out_unlock: | |
4ac91378 | 2037 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2038 | path_put(&nd.path); |
1da177e4 LT |
2039 | putname(tmp); |
2040 | ||
2041 | return error; | |
2042 | } | |
2043 | ||
3480b257 | 2044 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2045 | { |
2046 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2047 | } | |
2048 | ||
1da177e4 LT |
2049 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2050 | { | |
a95164d9 | 2051 | int error = may_create(dir, dentry); |
1da177e4 LT |
2052 | |
2053 | if (error) | |
2054 | return error; | |
2055 | ||
acfa4380 | 2056 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2057 | return -EPERM; |
2058 | ||
2059 | mode &= (S_IRWXUGO|S_ISVTX); | |
2060 | error = security_inode_mkdir(dir, dentry, mode); | |
2061 | if (error) | |
2062 | return error; | |
2063 | ||
1da177e4 | 2064 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2065 | if (!error) |
f38aa942 | 2066 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2067 | return error; |
2068 | } | |
2069 | ||
2e4d0924 | 2070 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2071 | { |
2072 | int error = 0; | |
2073 | char * tmp; | |
6902d925 DH |
2074 | struct dentry *dentry; |
2075 | struct nameidata nd; | |
1da177e4 | 2076 | |
2ad94ae6 AV |
2077 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2078 | if (error) | |
6902d925 | 2079 | goto out_err; |
1da177e4 | 2080 | |
6902d925 DH |
2081 | dentry = lookup_create(&nd, 1); |
2082 | error = PTR_ERR(dentry); | |
2083 | if (IS_ERR(dentry)) | |
2084 | goto out_unlock; | |
1da177e4 | 2085 | |
4ac91378 | 2086 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2087 | mode &= ~current_umask(); |
463c3197 DH |
2088 | error = mnt_want_write(nd.path.mnt); |
2089 | if (error) | |
2090 | goto out_dput; | |
be6d3e56 KT |
2091 | error = security_path_mkdir(&nd.path, dentry, mode); |
2092 | if (error) | |
2093 | goto out_drop_write; | |
4ac91378 | 2094 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2095 | out_drop_write: |
463c3197 DH |
2096 | mnt_drop_write(nd.path.mnt); |
2097 | out_dput: | |
6902d925 DH |
2098 | dput(dentry); |
2099 | out_unlock: | |
4ac91378 | 2100 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2101 | path_put(&nd.path); |
6902d925 DH |
2102 | putname(tmp); |
2103 | out_err: | |
1da177e4 LT |
2104 | return error; |
2105 | } | |
2106 | ||
3cdad428 | 2107 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2108 | { |
2109 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2110 | } | |
2111 | ||
1da177e4 LT |
2112 | /* |
2113 | * We try to drop the dentry early: we should have | |
2114 | * a usage count of 2 if we're the only user of this | |
2115 | * dentry, and if that is true (possibly after pruning | |
2116 | * the dcache), then we drop the dentry now. | |
2117 | * | |
2118 | * A low-level filesystem can, if it choses, legally | |
2119 | * do a | |
2120 | * | |
2121 | * if (!d_unhashed(dentry)) | |
2122 | * return -EBUSY; | |
2123 | * | |
2124 | * if it cannot handle the case of removing a directory | |
2125 | * that is still in use by something else.. | |
2126 | */ | |
2127 | void dentry_unhash(struct dentry *dentry) | |
2128 | { | |
2129 | dget(dentry); | |
dc168427 | 2130 | shrink_dcache_parent(dentry); |
1da177e4 | 2131 | spin_lock(&dentry->d_lock); |
b7ab39f6 | 2132 | if (dentry->d_count == 2) |
1da177e4 LT |
2133 | __d_drop(dentry); |
2134 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2135 | } |
2136 | ||
2137 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2138 | { | |
2139 | int error = may_delete(dir, dentry, 1); | |
2140 | ||
2141 | if (error) | |
2142 | return error; | |
2143 | ||
acfa4380 | 2144 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2145 | return -EPERM; |
2146 | ||
1b1dcc1b | 2147 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2148 | dentry_unhash(dentry); |
2149 | if (d_mountpoint(dentry)) | |
2150 | error = -EBUSY; | |
2151 | else { | |
2152 | error = security_inode_rmdir(dir, dentry); | |
2153 | if (!error) { | |
2154 | error = dir->i_op->rmdir(dir, dentry); | |
d83c49f3 | 2155 | if (!error) { |
1da177e4 | 2156 | dentry->d_inode->i_flags |= S_DEAD; |
d83c49f3 AV |
2157 | dont_mount(dentry); |
2158 | } | |
1da177e4 LT |
2159 | } |
2160 | } | |
1b1dcc1b | 2161 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2162 | if (!error) { |
1da177e4 LT |
2163 | d_delete(dentry); |
2164 | } | |
2165 | dput(dentry); | |
2166 | ||
2167 | return error; | |
2168 | } | |
2169 | ||
5590ff0d | 2170 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2171 | { |
2172 | int error = 0; | |
2173 | char * name; | |
2174 | struct dentry *dentry; | |
2175 | struct nameidata nd; | |
2176 | ||
2ad94ae6 | 2177 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2178 | if (error) |
2ad94ae6 | 2179 | return error; |
1da177e4 LT |
2180 | |
2181 | switch(nd.last_type) { | |
0612d9fb OH |
2182 | case LAST_DOTDOT: |
2183 | error = -ENOTEMPTY; | |
2184 | goto exit1; | |
2185 | case LAST_DOT: | |
2186 | error = -EINVAL; | |
2187 | goto exit1; | |
2188 | case LAST_ROOT: | |
2189 | error = -EBUSY; | |
2190 | goto exit1; | |
1da177e4 | 2191 | } |
0612d9fb OH |
2192 | |
2193 | nd.flags &= ~LOOKUP_PARENT; | |
2194 | ||
4ac91378 | 2195 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2196 | dentry = lookup_hash(&nd); |
1da177e4 | 2197 | error = PTR_ERR(dentry); |
6902d925 DH |
2198 | if (IS_ERR(dentry)) |
2199 | goto exit2; | |
0622753b DH |
2200 | error = mnt_want_write(nd.path.mnt); |
2201 | if (error) | |
2202 | goto exit3; | |
be6d3e56 KT |
2203 | error = security_path_rmdir(&nd.path, dentry); |
2204 | if (error) | |
2205 | goto exit4; | |
4ac91378 | 2206 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2207 | exit4: |
0622753b DH |
2208 | mnt_drop_write(nd.path.mnt); |
2209 | exit3: | |
6902d925 DH |
2210 | dput(dentry); |
2211 | exit2: | |
4ac91378 | 2212 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2213 | exit1: |
1d957f9b | 2214 | path_put(&nd.path); |
1da177e4 LT |
2215 | putname(name); |
2216 | return error; | |
2217 | } | |
2218 | ||
3cdad428 | 2219 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2220 | { |
2221 | return do_rmdir(AT_FDCWD, pathname); | |
2222 | } | |
2223 | ||
1da177e4 LT |
2224 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2225 | { | |
2226 | int error = may_delete(dir, dentry, 0); | |
2227 | ||
2228 | if (error) | |
2229 | return error; | |
2230 | ||
acfa4380 | 2231 | if (!dir->i_op->unlink) |
1da177e4 LT |
2232 | return -EPERM; |
2233 | ||
1b1dcc1b | 2234 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2235 | if (d_mountpoint(dentry)) |
2236 | error = -EBUSY; | |
2237 | else { | |
2238 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2239 | if (!error) { |
1da177e4 | 2240 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2241 | if (!error) |
d83c49f3 | 2242 | dont_mount(dentry); |
bec1052e | 2243 | } |
1da177e4 | 2244 | } |
1b1dcc1b | 2245 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2246 | |
2247 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2248 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2249 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2250 | d_delete(dentry); |
1da177e4 | 2251 | } |
0eeca283 | 2252 | |
1da177e4 LT |
2253 | return error; |
2254 | } | |
2255 | ||
2256 | /* | |
2257 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2258 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2259 | * writeout happening, and we don't want to prevent access to the directory |
2260 | * while waiting on the I/O. | |
2261 | */ | |
5590ff0d | 2262 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2263 | { |
2ad94ae6 AV |
2264 | int error; |
2265 | char *name; | |
1da177e4 LT |
2266 | struct dentry *dentry; |
2267 | struct nameidata nd; | |
2268 | struct inode *inode = NULL; | |
2269 | ||
2ad94ae6 | 2270 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2271 | if (error) |
2ad94ae6 AV |
2272 | return error; |
2273 | ||
1da177e4 LT |
2274 | error = -EISDIR; |
2275 | if (nd.last_type != LAST_NORM) | |
2276 | goto exit1; | |
0612d9fb OH |
2277 | |
2278 | nd.flags &= ~LOOKUP_PARENT; | |
2279 | ||
4ac91378 | 2280 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2281 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2282 | error = PTR_ERR(dentry); |
2283 | if (!IS_ERR(dentry)) { | |
2284 | /* Why not before? Because we want correct error value */ | |
2285 | if (nd.last.name[nd.last.len]) | |
2286 | goto slashes; | |
2287 | inode = dentry->d_inode; | |
2288 | if (inode) | |
7de9c6ee | 2289 | ihold(inode); |
0622753b DH |
2290 | error = mnt_want_write(nd.path.mnt); |
2291 | if (error) | |
2292 | goto exit2; | |
be6d3e56 KT |
2293 | error = security_path_unlink(&nd.path, dentry); |
2294 | if (error) | |
2295 | goto exit3; | |
4ac91378 | 2296 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2297 | exit3: |
0622753b | 2298 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2299 | exit2: |
2300 | dput(dentry); | |
2301 | } | |
4ac91378 | 2302 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2303 | if (inode) |
2304 | iput(inode); /* truncate the inode here */ | |
2305 | exit1: | |
1d957f9b | 2306 | path_put(&nd.path); |
1da177e4 LT |
2307 | putname(name); |
2308 | return error; | |
2309 | ||
2310 | slashes: | |
2311 | error = !dentry->d_inode ? -ENOENT : | |
2312 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2313 | goto exit2; | |
2314 | } | |
2315 | ||
2e4d0924 | 2316 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2317 | { |
2318 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2319 | return -EINVAL; | |
2320 | ||
2321 | if (flag & AT_REMOVEDIR) | |
2322 | return do_rmdir(dfd, pathname); | |
2323 | ||
2324 | return do_unlinkat(dfd, pathname); | |
2325 | } | |
2326 | ||
3480b257 | 2327 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2328 | { |
2329 | return do_unlinkat(AT_FDCWD, pathname); | |
2330 | } | |
2331 | ||
db2e747b | 2332 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2333 | { |
a95164d9 | 2334 | int error = may_create(dir, dentry); |
1da177e4 LT |
2335 | |
2336 | if (error) | |
2337 | return error; | |
2338 | ||
acfa4380 | 2339 | if (!dir->i_op->symlink) |
1da177e4 LT |
2340 | return -EPERM; |
2341 | ||
2342 | error = security_inode_symlink(dir, dentry, oldname); | |
2343 | if (error) | |
2344 | return error; | |
2345 | ||
1da177e4 | 2346 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2347 | if (!error) |
f38aa942 | 2348 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2349 | return error; |
2350 | } | |
2351 | ||
2e4d0924 HC |
2352 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2353 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2354 | { |
2ad94ae6 AV |
2355 | int error; |
2356 | char *from; | |
2357 | char *to; | |
6902d925 DH |
2358 | struct dentry *dentry; |
2359 | struct nameidata nd; | |
1da177e4 LT |
2360 | |
2361 | from = getname(oldname); | |
2ad94ae6 | 2362 | if (IS_ERR(from)) |
1da177e4 | 2363 | return PTR_ERR(from); |
1da177e4 | 2364 | |
2ad94ae6 | 2365 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2366 | if (error) |
2ad94ae6 AV |
2367 | goto out_putname; |
2368 | ||
6902d925 DH |
2369 | dentry = lookup_create(&nd, 0); |
2370 | error = PTR_ERR(dentry); | |
2371 | if (IS_ERR(dentry)) | |
2372 | goto out_unlock; | |
2373 | ||
75c3f29d DH |
2374 | error = mnt_want_write(nd.path.mnt); |
2375 | if (error) | |
2376 | goto out_dput; | |
be6d3e56 KT |
2377 | error = security_path_symlink(&nd.path, dentry, from); |
2378 | if (error) | |
2379 | goto out_drop_write; | |
db2e747b | 2380 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2381 | out_drop_write: |
75c3f29d DH |
2382 | mnt_drop_write(nd.path.mnt); |
2383 | out_dput: | |
6902d925 DH |
2384 | dput(dentry); |
2385 | out_unlock: | |
4ac91378 | 2386 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2387 | path_put(&nd.path); |
6902d925 DH |
2388 | putname(to); |
2389 | out_putname: | |
1da177e4 LT |
2390 | putname(from); |
2391 | return error; | |
2392 | } | |
2393 | ||
3480b257 | 2394 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2395 | { |
2396 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2397 | } | |
2398 | ||
1da177e4 LT |
2399 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2400 | { | |
2401 | struct inode *inode = old_dentry->d_inode; | |
2402 | int error; | |
2403 | ||
2404 | if (!inode) | |
2405 | return -ENOENT; | |
2406 | ||
a95164d9 | 2407 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2408 | if (error) |
2409 | return error; | |
2410 | ||
2411 | if (dir->i_sb != inode->i_sb) | |
2412 | return -EXDEV; | |
2413 | ||
2414 | /* | |
2415 | * A link to an append-only or immutable file cannot be created. | |
2416 | */ | |
2417 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2418 | return -EPERM; | |
acfa4380 | 2419 | if (!dir->i_op->link) |
1da177e4 | 2420 | return -EPERM; |
7e79eedb | 2421 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2422 | return -EPERM; |
2423 | ||
2424 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2425 | if (error) | |
2426 | return error; | |
2427 | ||
7e79eedb | 2428 | mutex_lock(&inode->i_mutex); |
1da177e4 | 2429 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2430 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2431 | if (!error) |
7e79eedb | 2432 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2433 | return error; |
2434 | } | |
2435 | ||
2436 | /* | |
2437 | * Hardlinks are often used in delicate situations. We avoid | |
2438 | * security-related surprises by not following symlinks on the | |
2439 | * newname. --KAB | |
2440 | * | |
2441 | * We don't follow them on the oldname either to be compatible | |
2442 | * with linux 2.0, and to avoid hard-linking to directories | |
2443 | * and other special files. --ADM | |
2444 | */ | |
2e4d0924 HC |
2445 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2446 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2447 | { |
2448 | struct dentry *new_dentry; | |
2d8f3038 AV |
2449 | struct nameidata nd; |
2450 | struct path old_path; | |
1da177e4 | 2451 | int error; |
2ad94ae6 | 2452 | char *to; |
1da177e4 | 2453 | |
45c9b11a | 2454 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2455 | return -EINVAL; |
2456 | ||
2d8f3038 AV |
2457 | error = user_path_at(olddfd, oldname, |
2458 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2459 | &old_path); | |
1da177e4 | 2460 | if (error) |
2ad94ae6 AV |
2461 | return error; |
2462 | ||
2463 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2464 | if (error) |
2465 | goto out; | |
2466 | error = -EXDEV; | |
2d8f3038 | 2467 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2468 | goto out_release; |
2469 | new_dentry = lookup_create(&nd, 0); | |
2470 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2471 | if (IS_ERR(new_dentry)) |
2472 | goto out_unlock; | |
75c3f29d DH |
2473 | error = mnt_want_write(nd.path.mnt); |
2474 | if (error) | |
2475 | goto out_dput; | |
be6d3e56 KT |
2476 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2477 | if (error) | |
2478 | goto out_drop_write; | |
2d8f3038 | 2479 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2480 | out_drop_write: |
75c3f29d DH |
2481 | mnt_drop_write(nd.path.mnt); |
2482 | out_dput: | |
6902d925 DH |
2483 | dput(new_dentry); |
2484 | out_unlock: | |
4ac91378 | 2485 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2486 | out_release: |
1d957f9b | 2487 | path_put(&nd.path); |
2ad94ae6 | 2488 | putname(to); |
1da177e4 | 2489 | out: |
2d8f3038 | 2490 | path_put(&old_path); |
1da177e4 LT |
2491 | |
2492 | return error; | |
2493 | } | |
2494 | ||
3480b257 | 2495 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2496 | { |
c04030e1 | 2497 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2498 | } |
2499 | ||
1da177e4 LT |
2500 | /* |
2501 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2502 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2503 | * Problems: | |
2504 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2505 | * b) race potential - two innocent renames can create a loop together. | |
2506 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2507 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2508 | * story. |
2509 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2510 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2511 | * whether the target exists). Solution: try to be smart with locking |
2512 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2513 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2514 | * move will be locked. Thus we can rank directories by the tree |
2515 | * (ancestors first) and rank all non-directories after them. | |
2516 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2517 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2518 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2519 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2520 | * we'd better make sure that there's no link(2) for them. | |
2521 | * d) some filesystems don't support opened-but-unlinked directories, | |
2522 | * either because of layout or because they are not ready to deal with | |
2523 | * all cases correctly. The latter will be fixed (taking this sort of | |
2524 | * stuff into VFS), but the former is not going away. Solution: the same | |
2525 | * trick as in rmdir(). | |
2526 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2527 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2528 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 2529 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
2530 | * locking]. |
2531 | */ | |
75c96f85 AB |
2532 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2533 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2534 | { |
2535 | int error = 0; | |
2536 | struct inode *target; | |
2537 | ||
2538 | /* | |
2539 | * If we are going to change the parent - check write permissions, | |
2540 | * we'll need to flip '..'. | |
2541 | */ | |
2542 | if (new_dir != old_dir) { | |
f419a2e3 | 2543 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2544 | if (error) |
2545 | return error; | |
2546 | } | |
2547 | ||
2548 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2549 | if (error) | |
2550 | return error; | |
2551 | ||
2552 | target = new_dentry->d_inode; | |
d83c49f3 | 2553 | if (target) |
1b1dcc1b | 2554 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2555 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2556 | error = -EBUSY; | |
d83c49f3 AV |
2557 | else { |
2558 | if (target) | |
2559 | dentry_unhash(new_dentry); | |
1da177e4 | 2560 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
d83c49f3 | 2561 | } |
1da177e4 | 2562 | if (target) { |
d83c49f3 | 2563 | if (!error) { |
1da177e4 | 2564 | target->i_flags |= S_DEAD; |
d83c49f3 AV |
2565 | dont_mount(new_dentry); |
2566 | } | |
1b1dcc1b | 2567 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2568 | if (d_unhashed(new_dentry)) |
2569 | d_rehash(new_dentry); | |
2570 | dput(new_dentry); | |
2571 | } | |
e31e14ec | 2572 | if (!error) |
349457cc MF |
2573 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2574 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2575 | return error; |
2576 | } | |
2577 | ||
75c96f85 AB |
2578 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2579 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2580 | { |
2581 | struct inode *target; | |
2582 | int error; | |
2583 | ||
2584 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2585 | if (error) | |
2586 | return error; | |
2587 | ||
2588 | dget(new_dentry); | |
2589 | target = new_dentry->d_inode; | |
2590 | if (target) | |
1b1dcc1b | 2591 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2592 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2593 | error = -EBUSY; | |
2594 | else | |
2595 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2596 | if (!error) { | |
bec1052e | 2597 | if (target) |
d83c49f3 | 2598 | dont_mount(new_dentry); |
349457cc | 2599 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2600 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2601 | } |
2602 | if (target) | |
1b1dcc1b | 2603 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2604 | dput(new_dentry); |
2605 | return error; | |
2606 | } | |
2607 | ||
2608 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2609 | struct inode *new_dir, struct dentry *new_dentry) | |
2610 | { | |
2611 | int error; | |
2612 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 2613 | const unsigned char *old_name; |
1da177e4 LT |
2614 | |
2615 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2616 | return 0; | |
2617 | ||
2618 | error = may_delete(old_dir, old_dentry, is_dir); | |
2619 | if (error) | |
2620 | return error; | |
2621 | ||
2622 | if (!new_dentry->d_inode) | |
a95164d9 | 2623 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2624 | else |
2625 | error = may_delete(new_dir, new_dentry, is_dir); | |
2626 | if (error) | |
2627 | return error; | |
2628 | ||
acfa4380 | 2629 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2630 | return -EPERM; |
2631 | ||
0eeca283 RL |
2632 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2633 | ||
1da177e4 LT |
2634 | if (is_dir) |
2635 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2636 | else | |
2637 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
2638 | if (!error) |
2639 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 2640 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
2641 | fsnotify_oldname_free(old_name); |
2642 | ||
1da177e4 LT |
2643 | return error; |
2644 | } | |
2645 | ||
2e4d0924 HC |
2646 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2647 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2648 | { |
2ad94ae6 AV |
2649 | struct dentry *old_dir, *new_dir; |
2650 | struct dentry *old_dentry, *new_dentry; | |
2651 | struct dentry *trap; | |
1da177e4 | 2652 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2653 | char *from; |
2654 | char *to; | |
2655 | int error; | |
1da177e4 | 2656 | |
2ad94ae6 | 2657 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2658 | if (error) |
2659 | goto exit; | |
2660 | ||
2ad94ae6 | 2661 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2662 | if (error) |
2663 | goto exit1; | |
2664 | ||
2665 | error = -EXDEV; | |
4ac91378 | 2666 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2667 | goto exit2; |
2668 | ||
4ac91378 | 2669 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2670 | error = -EBUSY; |
2671 | if (oldnd.last_type != LAST_NORM) | |
2672 | goto exit2; | |
2673 | ||
4ac91378 | 2674 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2675 | if (newnd.last_type != LAST_NORM) |
2676 | goto exit2; | |
2677 | ||
0612d9fb OH |
2678 | oldnd.flags &= ~LOOKUP_PARENT; |
2679 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2680 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2681 | |
1da177e4 LT |
2682 | trap = lock_rename(new_dir, old_dir); |
2683 | ||
49705b77 | 2684 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2685 | error = PTR_ERR(old_dentry); |
2686 | if (IS_ERR(old_dentry)) | |
2687 | goto exit3; | |
2688 | /* source must exist */ | |
2689 | error = -ENOENT; | |
2690 | if (!old_dentry->d_inode) | |
2691 | goto exit4; | |
2692 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2693 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2694 | error = -ENOTDIR; | |
2695 | if (oldnd.last.name[oldnd.last.len]) | |
2696 | goto exit4; | |
2697 | if (newnd.last.name[newnd.last.len]) | |
2698 | goto exit4; | |
2699 | } | |
2700 | /* source should not be ancestor of target */ | |
2701 | error = -EINVAL; | |
2702 | if (old_dentry == trap) | |
2703 | goto exit4; | |
49705b77 | 2704 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2705 | error = PTR_ERR(new_dentry); |
2706 | if (IS_ERR(new_dentry)) | |
2707 | goto exit4; | |
2708 | /* target should not be an ancestor of source */ | |
2709 | error = -ENOTEMPTY; | |
2710 | if (new_dentry == trap) | |
2711 | goto exit5; | |
2712 | ||
9079b1eb DH |
2713 | error = mnt_want_write(oldnd.path.mnt); |
2714 | if (error) | |
2715 | goto exit5; | |
be6d3e56 KT |
2716 | error = security_path_rename(&oldnd.path, old_dentry, |
2717 | &newnd.path, new_dentry); | |
2718 | if (error) | |
2719 | goto exit6; | |
1da177e4 LT |
2720 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2721 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2722 | exit6: |
9079b1eb | 2723 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2724 | exit5: |
2725 | dput(new_dentry); | |
2726 | exit4: | |
2727 | dput(old_dentry); | |
2728 | exit3: | |
2729 | unlock_rename(new_dir, old_dir); | |
2730 | exit2: | |
1d957f9b | 2731 | path_put(&newnd.path); |
2ad94ae6 | 2732 | putname(to); |
1da177e4 | 2733 | exit1: |
1d957f9b | 2734 | path_put(&oldnd.path); |
1da177e4 | 2735 | putname(from); |
2ad94ae6 | 2736 | exit: |
1da177e4 LT |
2737 | return error; |
2738 | } | |
2739 | ||
a26eab24 | 2740 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2741 | { |
2742 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2743 | } | |
2744 | ||
1da177e4 LT |
2745 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2746 | { | |
2747 | int len; | |
2748 | ||
2749 | len = PTR_ERR(link); | |
2750 | if (IS_ERR(link)) | |
2751 | goto out; | |
2752 | ||
2753 | len = strlen(link); | |
2754 | if (len > (unsigned) buflen) | |
2755 | len = buflen; | |
2756 | if (copy_to_user(buffer, link, len)) | |
2757 | len = -EFAULT; | |
2758 | out: | |
2759 | return len; | |
2760 | } | |
2761 | ||
2762 | /* | |
2763 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2764 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2765 | * using) it for any given inode is up to filesystem. | |
2766 | */ | |
2767 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2768 | { | |
2769 | struct nameidata nd; | |
cc314eef | 2770 | void *cookie; |
694a1764 | 2771 | int res; |
cc314eef | 2772 | |
1da177e4 | 2773 | nd.depth = 0; |
cc314eef | 2774 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2775 | if (IS_ERR(cookie)) |
2776 | return PTR_ERR(cookie); | |
2777 | ||
2778 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2779 | if (dentry->d_inode->i_op->put_link) | |
2780 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2781 | return res; | |
1da177e4 LT |
2782 | } |
2783 | ||
2784 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2785 | { | |
2786 | return __vfs_follow_link(nd, link); | |
2787 | } | |
2788 | ||
2789 | /* get the link contents into pagecache */ | |
2790 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2791 | { | |
ebd09abb DG |
2792 | char *kaddr; |
2793 | struct page *page; | |
1da177e4 | 2794 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2795 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2796 | if (IS_ERR(page)) |
6fe6900e | 2797 | return (char*)page; |
1da177e4 | 2798 | *ppage = page; |
ebd09abb DG |
2799 | kaddr = kmap(page); |
2800 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2801 | return kaddr; | |
1da177e4 LT |
2802 | } |
2803 | ||
2804 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2805 | { | |
2806 | struct page *page = NULL; | |
2807 | char *s = page_getlink(dentry, &page); | |
2808 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2809 | if (page) { | |
2810 | kunmap(page); | |
2811 | page_cache_release(page); | |
2812 | } | |
2813 | return res; | |
2814 | } | |
2815 | ||
cc314eef | 2816 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2817 | { |
cc314eef | 2818 | struct page *page = NULL; |
1da177e4 | 2819 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2820 | return page; |
1da177e4 LT |
2821 | } |
2822 | ||
cc314eef | 2823 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2824 | { |
cc314eef LT |
2825 | struct page *page = cookie; |
2826 | ||
2827 | if (page) { | |
1da177e4 LT |
2828 | kunmap(page); |
2829 | page_cache_release(page); | |
1da177e4 LT |
2830 | } |
2831 | } | |
2832 | ||
54566b2c NP |
2833 | /* |
2834 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2835 | */ | |
2836 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2837 | { |
2838 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2839 | struct page *page; |
afddba49 | 2840 | void *fsdata; |
beb497ab | 2841 | int err; |
1da177e4 | 2842 | char *kaddr; |
54566b2c NP |
2843 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2844 | if (nofs) | |
2845 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2846 | |
7e53cac4 | 2847 | retry: |
afddba49 | 2848 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2849 | flags, &page, &fsdata); |
1da177e4 | 2850 | if (err) |
afddba49 NP |
2851 | goto fail; |
2852 | ||
1da177e4 LT |
2853 | kaddr = kmap_atomic(page, KM_USER0); |
2854 | memcpy(kaddr, symname, len-1); | |
2855 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2856 | |
2857 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2858 | page, fsdata); | |
1da177e4 LT |
2859 | if (err < 0) |
2860 | goto fail; | |
afddba49 NP |
2861 | if (err < len-1) |
2862 | goto retry; | |
2863 | ||
1da177e4 LT |
2864 | mark_inode_dirty(inode); |
2865 | return 0; | |
1da177e4 LT |
2866 | fail: |
2867 | return err; | |
2868 | } | |
2869 | ||
0adb25d2 KK |
2870 | int page_symlink(struct inode *inode, const char *symname, int len) |
2871 | { | |
2872 | return __page_symlink(inode, symname, len, | |
54566b2c | 2873 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2874 | } |
2875 | ||
92e1d5be | 2876 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2877 | .readlink = generic_readlink, |
2878 | .follow_link = page_follow_link_light, | |
2879 | .put_link = page_put_link, | |
2880 | }; | |
2881 | ||
2d8f3038 | 2882 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2883 | EXPORT_SYMBOL(follow_down); |
2884 | EXPORT_SYMBOL(follow_up); | |
2885 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2886 | EXPORT_SYMBOL(getname); | |
2887 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2888 | EXPORT_SYMBOL(lookup_one_len); |
2889 | EXPORT_SYMBOL(page_follow_link_light); | |
2890 | EXPORT_SYMBOL(page_put_link); | |
2891 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2892 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2893 | EXPORT_SYMBOL(page_symlink); |
2894 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2895 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2896 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2897 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2898 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2899 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2900 | EXPORT_SYMBOL(unlock_rename); |
2901 | EXPORT_SYMBOL(vfs_create); | |
2902 | EXPORT_SYMBOL(vfs_follow_link); | |
2903 | EXPORT_SYMBOL(vfs_link); | |
2904 | EXPORT_SYMBOL(vfs_mkdir); | |
2905 | EXPORT_SYMBOL(vfs_mknod); | |
2906 | EXPORT_SYMBOL(generic_permission); | |
2907 | EXPORT_SYMBOL(vfs_readlink); | |
2908 | EXPORT_SYMBOL(vfs_rename); | |
2909 | EXPORT_SYMBOL(vfs_rmdir); | |
2910 | EXPORT_SYMBOL(vfs_symlink); | |
2911 | EXPORT_SYMBOL(vfs_unlink); | |
2912 | EXPORT_SYMBOL(dentry_unhash); | |
2913 | EXPORT_SYMBOL(generic_readlink); |