]>
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 | ||
615 | /* no need for dcache_lock, as serialization is taken care in | |
616 | * namespace.c | |
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 | ||
648 | /* no need for dcache_lock, as serialization is taken care in | |
649 | * namespace.c | |
650 | */ | |
9393bd07 | 651 | int follow_down(struct path *path) |
1da177e4 LT |
652 | { |
653 | struct vfsmount *mounted; | |
654 | ||
1c755af4 | 655 | mounted = lookup_mnt(path); |
1da177e4 | 656 | if (mounted) { |
9393bd07 AV |
657 | dput(path->dentry); |
658 | mntput(path->mnt); | |
659 | path->mnt = mounted; | |
660 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
661 | return 1; |
662 | } | |
663 | return 0; | |
664 | } | |
665 | ||
f1662356 | 666 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 667 | { |
2a737871 | 668 | set_root(nd); |
e518ddb7 | 669 | |
1da177e4 | 670 | while(1) { |
4ac91378 | 671 | struct dentry *old = nd->path.dentry; |
1da177e4 | 672 | |
2a737871 AV |
673 | if (nd->path.dentry == nd->root.dentry && |
674 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
675 | break; |
676 | } | |
4ac91378 | 677 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
678 | /* rare case of legitimate dget_parent()... */ |
679 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
680 | dput(old); |
681 | break; | |
682 | } | |
3088dd70 | 683 | if (!follow_up(&nd->path)) |
1da177e4 | 684 | break; |
1da177e4 | 685 | } |
79ed0226 | 686 | follow_mount(&nd->path); |
1da177e4 LT |
687 | } |
688 | ||
baa03890 NP |
689 | /* |
690 | * Allocate a dentry with name and parent, and perform a parent | |
691 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
692 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
693 | * have verified that no child exists while under i_mutex. | |
694 | */ | |
695 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
696 | struct qstr *name, struct nameidata *nd) | |
697 | { | |
698 | struct inode *inode = parent->d_inode; | |
699 | struct dentry *dentry; | |
700 | struct dentry *old; | |
701 | ||
702 | /* Don't create child dentry for a dead directory. */ | |
703 | if (unlikely(IS_DEADDIR(inode))) | |
704 | return ERR_PTR(-ENOENT); | |
705 | ||
706 | dentry = d_alloc(parent, name); | |
707 | if (unlikely(!dentry)) | |
708 | return ERR_PTR(-ENOMEM); | |
709 | ||
710 | old = inode->i_op->lookup(inode, dentry, nd); | |
711 | if (unlikely(old)) { | |
712 | dput(dentry); | |
713 | dentry = old; | |
714 | } | |
715 | return dentry; | |
716 | } | |
717 | ||
1da177e4 LT |
718 | /* |
719 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
720 | * small and for now I'd prefer to have fast path as straight as possible. | |
721 | * It _is_ time-critical. | |
722 | */ | |
723 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
724 | struct path *path) | |
725 | { | |
4ac91378 | 726 | struct vfsmount *mnt = nd->path.mnt; |
6e6b1bd1 AV |
727 | struct dentry *dentry, *parent; |
728 | struct inode *dir; | |
3cac260a AV |
729 | /* |
730 | * See if the low-level filesystem might want | |
731 | * to use its own hash.. | |
732 | */ | |
733 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { | |
734 | int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, name); | |
735 | if (err < 0) | |
736 | return err; | |
737 | } | |
1da177e4 | 738 | |
b04f784e NP |
739 | /* |
740 | * Rename seqlock is not required here because in the off chance | |
741 | * of a false negative due to a concurrent rename, we're going to | |
742 | * do the non-racy lookup, below. | |
743 | */ | |
3cac260a | 744 | dentry = __d_lookup(nd->path.dentry, name); |
1da177e4 LT |
745 | if (!dentry) |
746 | goto need_lookup; | |
2e2e88ea | 747 | found: |
1da177e4 LT |
748 | if (dentry->d_op && dentry->d_op->d_revalidate) |
749 | goto need_revalidate; | |
750 | done: | |
751 | path->mnt = mnt; | |
752 | path->dentry = dentry; | |
634ee701 | 753 | __follow_mount(path); |
1da177e4 LT |
754 | return 0; |
755 | ||
756 | need_lookup: | |
6e6b1bd1 AV |
757 | parent = nd->path.dentry; |
758 | dir = parent->d_inode; | |
759 | ||
760 | mutex_lock(&dir->i_mutex); | |
761 | /* | |
762 | * First re-do the cached lookup just in case it was created | |
b04f784e NP |
763 | * while we waited for the directory semaphore, or the first |
764 | * lookup failed due to an unrelated rename. | |
6e6b1bd1 | 765 | * |
b04f784e NP |
766 | * This could use version numbering or similar to avoid unnecessary |
767 | * cache lookups, but then we'd have to do the first lookup in the | |
768 | * non-racy way. However in the common case here, everything should | |
769 | * be hot in cache, so would it be a big win? | |
6e6b1bd1 AV |
770 | */ |
771 | dentry = d_lookup(parent, name); | |
baa03890 NP |
772 | if (likely(!dentry)) { |
773 | dentry = d_alloc_and_lookup(parent, name, nd); | |
6e6b1bd1 AV |
774 | mutex_unlock(&dir->i_mutex); |
775 | if (IS_ERR(dentry)) | |
776 | goto fail; | |
777 | goto done; | |
778 | } | |
6e6b1bd1 AV |
779 | /* |
780 | * Uhhuh! Nasty case: the cache was re-populated while | |
781 | * we waited on the semaphore. Need to revalidate. | |
782 | */ | |
783 | mutex_unlock(&dir->i_mutex); | |
2e2e88ea | 784 | goto found; |
1da177e4 LT |
785 | |
786 | need_revalidate: | |
bcdc5e01 IK |
787 | dentry = do_revalidate(dentry, nd); |
788 | if (!dentry) | |
789 | goto need_lookup; | |
790 | if (IS_ERR(dentry)) | |
791 | goto fail; | |
792 | goto done; | |
1da177e4 LT |
793 | |
794 | fail: | |
795 | return PTR_ERR(dentry); | |
796 | } | |
797 | ||
ac278a9c AV |
798 | /* |
799 | * This is a temporary kludge to deal with "automount" symlinks; proper | |
800 | * solution is to trigger them on follow_mount(), so that do_lookup() | |
801 | * would DTRT. To be killed before 2.6.34-final. | |
802 | */ | |
803 | static inline int follow_on_final(struct inode *inode, unsigned lookup_flags) | |
804 | { | |
805 | return inode && unlikely(inode->i_op->follow_link) && | |
806 | ((lookup_flags & LOOKUP_FOLLOW) || S_ISDIR(inode->i_mode)); | |
807 | } | |
808 | ||
1da177e4 LT |
809 | /* |
810 | * Name resolution. | |
ea3834d9 PM |
811 | * This is the basic name resolution function, turning a pathname into |
812 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 813 | * |
ea3834d9 PM |
814 | * Returns 0 and nd will have valid dentry and mnt on success. |
815 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 816 | */ |
6de88d72 | 817 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
818 | { |
819 | struct path next; | |
820 | struct inode *inode; | |
821 | int err; | |
822 | unsigned int lookup_flags = nd->flags; | |
823 | ||
824 | while (*name=='/') | |
825 | name++; | |
826 | if (!*name) | |
827 | goto return_reval; | |
828 | ||
4ac91378 | 829 | inode = nd->path.dentry->d_inode; |
1da177e4 | 830 | if (nd->depth) |
f55eab82 | 831 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
832 | |
833 | /* At this point we know we have a real path component. */ | |
834 | for(;;) { | |
835 | unsigned long hash; | |
836 | struct qstr this; | |
837 | unsigned int c; | |
838 | ||
cdce5d6b | 839 | nd->flags |= LOOKUP_CONTINUE; |
b75b5086 | 840 | err = exec_permission(inode); |
1da177e4 LT |
841 | if (err) |
842 | break; | |
843 | ||
844 | this.name = name; | |
845 | c = *(const unsigned char *)name; | |
846 | ||
847 | hash = init_name_hash(); | |
848 | do { | |
849 | name++; | |
850 | hash = partial_name_hash(c, hash); | |
851 | c = *(const unsigned char *)name; | |
852 | } while (c && (c != '/')); | |
853 | this.len = name - (const char *) this.name; | |
854 | this.hash = end_name_hash(hash); | |
855 | ||
856 | /* remove trailing slashes? */ | |
857 | if (!c) | |
858 | goto last_component; | |
859 | while (*++name == '/'); | |
860 | if (!*name) | |
861 | goto last_with_slashes; | |
862 | ||
863 | /* | |
864 | * "." and ".." are special - ".." especially so because it has | |
865 | * to be able to know about the current root directory and | |
866 | * parent relationships. | |
867 | */ | |
868 | if (this.name[0] == '.') switch (this.len) { | |
869 | default: | |
870 | break; | |
871 | case 2: | |
872 | if (this.name[1] != '.') | |
873 | break; | |
58c465eb | 874 | follow_dotdot(nd); |
4ac91378 | 875 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
876 | /* fallthrough */ |
877 | case 1: | |
878 | continue; | |
879 | } | |
1da177e4 LT |
880 | /* This does the actual lookups.. */ |
881 | err = do_lookup(nd, &this, &next); | |
882 | if (err) | |
883 | break; | |
1da177e4 LT |
884 | |
885 | err = -ENOENT; | |
886 | inode = next.dentry->d_inode; | |
887 | if (!inode) | |
888 | goto out_dput; | |
1da177e4 LT |
889 | |
890 | if (inode->i_op->follow_link) { | |
90ebe565 | 891 | err = do_follow_link(&next, nd); |
1da177e4 LT |
892 | if (err) |
893 | goto return_err; | |
894 | err = -ENOENT; | |
4ac91378 | 895 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
896 | if (!inode) |
897 | break; | |
09dd17d3 MS |
898 | } else |
899 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
900 | err = -ENOTDIR; |
901 | if (!inode->i_op->lookup) | |
902 | break; | |
903 | continue; | |
904 | /* here ends the main loop */ | |
905 | ||
906 | last_with_slashes: | |
907 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
908 | last_component: | |
f55eab82 TM |
909 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
910 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
911 | if (lookup_flags & LOOKUP_PARENT) |
912 | goto lookup_parent; | |
913 | if (this.name[0] == '.') switch (this.len) { | |
914 | default: | |
915 | break; | |
916 | case 2: | |
917 | if (this.name[1] != '.') | |
918 | break; | |
58c465eb | 919 | follow_dotdot(nd); |
4ac91378 | 920 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
921 | /* fallthrough */ |
922 | case 1: | |
923 | goto return_reval; | |
924 | } | |
1da177e4 LT |
925 | err = do_lookup(nd, &this, &next); |
926 | if (err) | |
927 | break; | |
1da177e4 | 928 | inode = next.dentry->d_inode; |
ac278a9c | 929 | if (follow_on_final(inode, lookup_flags)) { |
90ebe565 | 930 | err = do_follow_link(&next, nd); |
1da177e4 LT |
931 | if (err) |
932 | goto return_err; | |
4ac91378 | 933 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
934 | } else |
935 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
936 | err = -ENOENT; |
937 | if (!inode) | |
938 | break; | |
939 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
940 | err = -ENOTDIR; | |
acfa4380 | 941 | if (!inode->i_op->lookup) |
1da177e4 LT |
942 | break; |
943 | } | |
944 | goto return_base; | |
945 | lookup_parent: | |
946 | nd->last = this; | |
947 | nd->last_type = LAST_NORM; | |
948 | if (this.name[0] != '.') | |
949 | goto return_base; | |
950 | if (this.len == 1) | |
951 | nd->last_type = LAST_DOT; | |
952 | else if (this.len == 2 && this.name[1] == '.') | |
953 | nd->last_type = LAST_DOTDOT; | |
954 | else | |
955 | goto return_base; | |
956 | return_reval: | |
957 | /* | |
958 | * We bypassed the ordinary revalidation routines. | |
959 | * We may need to check the cached dentry for staleness. | |
960 | */ | |
4ac91378 JB |
961 | if (nd->path.dentry && nd->path.dentry->d_sb && |
962 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
963 | err = -ESTALE; |
964 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
965 | if (!nd->path.dentry->d_op->d_revalidate( |
966 | nd->path.dentry, nd)) | |
1da177e4 LT |
967 | break; |
968 | } | |
969 | return_base: | |
970 | return 0; | |
971 | out_dput: | |
1d957f9b | 972 | path_put_conditional(&next, nd); |
1da177e4 LT |
973 | break; |
974 | } | |
1d957f9b | 975 | path_put(&nd->path); |
1da177e4 LT |
976 | return_err: |
977 | return err; | |
978 | } | |
979 | ||
fc9b52cd | 980 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 981 | { |
6de88d72 AV |
982 | struct path save = nd->path; |
983 | int result; | |
984 | ||
1da177e4 | 985 | current->total_link_count = 0; |
6de88d72 AV |
986 | |
987 | /* make sure the stuff we saved doesn't go away */ | |
988 | path_get(&save); | |
989 | ||
990 | result = link_path_walk(name, nd); | |
991 | if (result == -ESTALE) { | |
992 | /* nd->path had been dropped */ | |
993 | current->total_link_count = 0; | |
994 | nd->path = save; | |
995 | path_get(&nd->path); | |
996 | nd->flags |= LOOKUP_REVAL; | |
997 | result = link_path_walk(name, nd); | |
998 | } | |
999 | ||
1000 | path_put(&save); | |
1001 | ||
1002 | return result; | |
1da177e4 LT |
1003 | } |
1004 | ||
9b4a9b14 | 1005 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1006 | { |
ea3834d9 | 1007 | int retval = 0; |
170aa3d0 UD |
1008 | int fput_needed; |
1009 | struct file *file; | |
1da177e4 LT |
1010 | |
1011 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1012 | nd->flags = flags; | |
1013 | nd->depth = 0; | |
2a737871 | 1014 | nd->root.mnt = NULL; |
1da177e4 | 1015 | |
1da177e4 | 1016 | if (*name=='/') { |
2a737871 AV |
1017 | set_root(nd); |
1018 | nd->path = nd->root; | |
1019 | path_get(&nd->root); | |
5590ff0d | 1020 | } else if (dfd == AT_FDCWD) { |
f7ad3c6b | 1021 | get_fs_pwd(current->fs, &nd->path); |
5590ff0d | 1022 | } else { |
5590ff0d UD |
1023 | struct dentry *dentry; |
1024 | ||
1025 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1026 | retval = -EBADF; |
1027 | if (!file) | |
6d09bb62 | 1028 | goto out_fail; |
5590ff0d | 1029 | |
0f7fc9e4 | 1030 | dentry = file->f_path.dentry; |
5590ff0d | 1031 | |
170aa3d0 UD |
1032 | retval = -ENOTDIR; |
1033 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1034 | goto fput_fail; |
5590ff0d UD |
1035 | |
1036 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1037 | if (retval) |
6d09bb62 | 1038 | goto fput_fail; |
5590ff0d | 1039 | |
5dd784d0 JB |
1040 | nd->path = file->f_path; |
1041 | path_get(&file->f_path); | |
5590ff0d UD |
1042 | |
1043 | fput_light(file, fput_needed); | |
1da177e4 | 1044 | } |
9b4a9b14 | 1045 | return 0; |
2dfdd266 | 1046 | |
9b4a9b14 AV |
1047 | fput_fail: |
1048 | fput_light(file, fput_needed); | |
1049 | out_fail: | |
1050 | return retval; | |
1051 | } | |
1052 | ||
1053 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1054 | static int do_path_lookup(int dfd, const char *name, | |
1055 | unsigned int flags, struct nameidata *nd) | |
1056 | { | |
1057 | int retval = path_init(dfd, name, flags, nd); | |
1058 | if (!retval) | |
1059 | retval = path_walk(name, nd); | |
4ac91378 JB |
1060 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1061 | nd->path.dentry->d_inode)) | |
1062 | audit_inode(name, nd->path.dentry); | |
2a737871 AV |
1063 | if (nd->root.mnt) { |
1064 | path_put(&nd->root); | |
1065 | nd->root.mnt = NULL; | |
1066 | } | |
170aa3d0 | 1067 | return retval; |
1da177e4 LT |
1068 | } |
1069 | ||
fc9b52cd | 1070 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1071 | struct nameidata *nd) |
1072 | { | |
1073 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1074 | } | |
1075 | ||
d1811465 AV |
1076 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1077 | { | |
1078 | struct nameidata nd; | |
1079 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1080 | if (!res) | |
1081 | *path = nd.path; | |
1082 | return res; | |
1083 | } | |
1084 | ||
16f18200 JJS |
1085 | /** |
1086 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1087 | * @dentry: pointer to dentry of the base directory | |
1088 | * @mnt: pointer to vfs mount of the base directory | |
1089 | * @name: pointer to file name | |
1090 | * @flags: lookup flags | |
1091 | * @nd: pointer to nameidata | |
1092 | */ | |
1093 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1094 | const char *name, unsigned int flags, | |
1095 | struct nameidata *nd) | |
1096 | { | |
1097 | int retval; | |
1098 | ||
1099 | /* same as do_path_lookup */ | |
1100 | nd->last_type = LAST_ROOT; | |
1101 | nd->flags = flags; | |
1102 | nd->depth = 0; | |
1103 | ||
c8e7f449 JB |
1104 | nd->path.dentry = dentry; |
1105 | nd->path.mnt = mnt; | |
1106 | path_get(&nd->path); | |
5b857119 AV |
1107 | nd->root = nd->path; |
1108 | path_get(&nd->root); | |
16f18200 JJS |
1109 | |
1110 | retval = path_walk(name, nd); | |
4ac91378 JB |
1111 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1112 | nd->path.dentry->d_inode)) | |
1113 | audit_inode(name, nd->path.dentry); | |
16f18200 | 1114 | |
5b857119 AV |
1115 | path_put(&nd->root); |
1116 | nd->root.mnt = NULL; | |
16f18200 | 1117 | |
2a737871 | 1118 | return retval; |
16f18200 JJS |
1119 | } |
1120 | ||
eead1911 CH |
1121 | static struct dentry *__lookup_hash(struct qstr *name, |
1122 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1123 | { |
057f6c01 | 1124 | struct dentry *dentry; |
1da177e4 LT |
1125 | struct inode *inode; |
1126 | int err; | |
1127 | ||
1128 | inode = base->d_inode; | |
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) { | |
1135 | err = base->d_op->d_hash(base, name); | |
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 | { |
057f6c01 JM |
1164 | int err; |
1165 | ||
b75b5086 | 1166 | err = exec_permission(nd->path.dentry->d_inode); |
057f6c01 | 1167 | if (err) |
eead1911 | 1168 | return ERR_PTR(err); |
4ac91378 | 1169 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1170 | } |
1171 | ||
eead1911 CH |
1172 | static int __lookup_one_len(const char *name, struct qstr *this, |
1173 | struct dentry *base, int len) | |
1da177e4 LT |
1174 | { |
1175 | unsigned long hash; | |
1da177e4 LT |
1176 | unsigned int c; |
1177 | ||
057f6c01 JM |
1178 | this->name = name; |
1179 | this->len = len; | |
1da177e4 | 1180 | if (!len) |
057f6c01 | 1181 | return -EACCES; |
1da177e4 LT |
1182 | |
1183 | hash = init_name_hash(); | |
1184 | while (len--) { | |
1185 | c = *(const unsigned char *)name++; | |
1186 | if (c == '/' || c == '\0') | |
057f6c01 | 1187 | return -EACCES; |
1da177e4 LT |
1188 | hash = partial_name_hash(c, hash); |
1189 | } | |
057f6c01 JM |
1190 | this->hash = end_name_hash(hash); |
1191 | return 0; | |
1192 | } | |
1da177e4 | 1193 | |
eead1911 | 1194 | /** |
a6b91919 | 1195 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1196 | * @name: pathname component to lookup |
1197 | * @base: base directory to lookup from | |
1198 | * @len: maximum length @len should be interpreted to | |
1199 | * | |
a6b91919 RD |
1200 | * Note that this routine is purely a helper for filesystem usage and should |
1201 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1202 | * nameidata argument is passed to the filesystem methods and a filesystem |
1203 | * using this helper needs to be prepared for that. | |
1204 | */ | |
057f6c01 JM |
1205 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1206 | { | |
1207 | int err; | |
1208 | struct qstr this; | |
1209 | ||
2f9092e1 DW |
1210 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1211 | ||
057f6c01 | 1212 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1213 | if (err) |
1214 | return ERR_PTR(err); | |
1215 | ||
b75b5086 | 1216 | err = exec_permission(base->d_inode); |
057f6c01 JM |
1217 | if (err) |
1218 | return ERR_PTR(err); | |
49705b77 | 1219 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1220 | } |
1221 | ||
2d8f3038 AV |
1222 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1223 | struct path *path) | |
1da177e4 | 1224 | { |
2d8f3038 | 1225 | struct nameidata nd; |
1da177e4 LT |
1226 | char *tmp = getname(name); |
1227 | int err = PTR_ERR(tmp); | |
1da177e4 | 1228 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1229 | |
1230 | BUG_ON(flags & LOOKUP_PARENT); | |
1231 | ||
1232 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1233 | putname(tmp); |
2d8f3038 AV |
1234 | if (!err) |
1235 | *path = nd.path; | |
1da177e4 LT |
1236 | } |
1237 | return err; | |
1238 | } | |
1239 | ||
2ad94ae6 AV |
1240 | static int user_path_parent(int dfd, const char __user *path, |
1241 | struct nameidata *nd, char **name) | |
1242 | { | |
1243 | char *s = getname(path); | |
1244 | int error; | |
1245 | ||
1246 | if (IS_ERR(s)) | |
1247 | return PTR_ERR(s); | |
1248 | ||
1249 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1250 | if (error) | |
1251 | putname(s); | |
1252 | else | |
1253 | *name = s; | |
1254 | ||
1255 | return error; | |
1256 | } | |
1257 | ||
1da177e4 LT |
1258 | /* |
1259 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1260 | * minimal. | |
1261 | */ | |
1262 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1263 | { | |
da9592ed DH |
1264 | uid_t fsuid = current_fsuid(); |
1265 | ||
1da177e4 LT |
1266 | if (!(dir->i_mode & S_ISVTX)) |
1267 | return 0; | |
da9592ed | 1268 | if (inode->i_uid == fsuid) |
1da177e4 | 1269 | return 0; |
da9592ed | 1270 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1271 | return 0; |
1272 | return !capable(CAP_FOWNER); | |
1273 | } | |
1274 | ||
1275 | /* | |
1276 | * Check whether we can remove a link victim from directory dir, check | |
1277 | * whether the type of victim is right. | |
1278 | * 1. We can't do it if dir is read-only (done in permission()) | |
1279 | * 2. We should have write and exec permissions on dir | |
1280 | * 3. We can't remove anything from append-only dir | |
1281 | * 4. We can't do anything with immutable dir (done in permission()) | |
1282 | * 5. If the sticky bit on dir is set we should either | |
1283 | * a. be owner of dir, or | |
1284 | * b. be owner of victim, or | |
1285 | * c. have CAP_FOWNER capability | |
1286 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1287 | * links pointing to it. | |
1288 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1289 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1290 | * 9. We can't remove a root or mountpoint. | |
1291 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1292 | * nfs_async_unlink(). | |
1293 | */ | |
858119e1 | 1294 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1295 | { |
1296 | int error; | |
1297 | ||
1298 | if (!victim->d_inode) | |
1299 | return -ENOENT; | |
1300 | ||
1301 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1302 | audit_inode_child(victim, dir); |
1da177e4 | 1303 | |
f419a2e3 | 1304 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1305 | if (error) |
1306 | return error; | |
1307 | if (IS_APPEND(dir)) | |
1308 | return -EPERM; | |
1309 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1310 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1311 | return -EPERM; |
1312 | if (isdir) { | |
1313 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1314 | return -ENOTDIR; | |
1315 | if (IS_ROOT(victim)) | |
1316 | return -EBUSY; | |
1317 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1318 | return -EISDIR; | |
1319 | if (IS_DEADDIR(dir)) | |
1320 | return -ENOENT; | |
1321 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1322 | return -EBUSY; | |
1323 | return 0; | |
1324 | } | |
1325 | ||
1326 | /* Check whether we can create an object with dentry child in directory | |
1327 | * dir. | |
1328 | * 1. We can't do it if child already exists (open has special treatment for | |
1329 | * this case, but since we are inlined it's OK) | |
1330 | * 2. We can't do it if dir is read-only (done in permission()) | |
1331 | * 3. We should have write and exec permissions on dir | |
1332 | * 4. We can't do it if dir is immutable (done in permission()) | |
1333 | */ | |
a95164d9 | 1334 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1335 | { |
1336 | if (child->d_inode) | |
1337 | return -EEXIST; | |
1338 | if (IS_DEADDIR(dir)) | |
1339 | return -ENOENT; | |
f419a2e3 | 1340 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1341 | } |
1342 | ||
1da177e4 LT |
1343 | /* |
1344 | * p1 and p2 should be directories on the same fs. | |
1345 | */ | |
1346 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1347 | { | |
1348 | struct dentry *p; | |
1349 | ||
1350 | if (p1 == p2) { | |
f2eace23 | 1351 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1352 | return NULL; |
1353 | } | |
1354 | ||
a11f3a05 | 1355 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1356 | |
e2761a11 OH |
1357 | p = d_ancestor(p2, p1); |
1358 | if (p) { | |
1359 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1360 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1361 | return p; | |
1da177e4 LT |
1362 | } |
1363 | ||
e2761a11 OH |
1364 | p = d_ancestor(p1, p2); |
1365 | if (p) { | |
1366 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1367 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1368 | return p; | |
1da177e4 LT |
1369 | } |
1370 | ||
f2eace23 IM |
1371 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1372 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1373 | return NULL; |
1374 | } | |
1375 | ||
1376 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1377 | { | |
1b1dcc1b | 1378 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1379 | if (p1 != p2) { |
1b1dcc1b | 1380 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1381 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1382 | } |
1383 | } | |
1384 | ||
1385 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1386 | struct nameidata *nd) | |
1387 | { | |
a95164d9 | 1388 | int error = may_create(dir, dentry); |
1da177e4 LT |
1389 | |
1390 | if (error) | |
1391 | return error; | |
1392 | ||
acfa4380 | 1393 | if (!dir->i_op->create) |
1da177e4 LT |
1394 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1395 | mode &= S_IALLUGO; | |
1396 | mode |= S_IFREG; | |
1397 | error = security_inode_create(dir, dentry, mode); | |
1398 | if (error) | |
1399 | return error; | |
1da177e4 | 1400 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1401 | if (!error) |
f38aa942 | 1402 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1403 | return error; |
1404 | } | |
1405 | ||
3fb64190 | 1406 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1407 | { |
3fb64190 | 1408 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1409 | struct inode *inode = dentry->d_inode; |
1410 | int error; | |
1411 | ||
1412 | if (!inode) | |
1413 | return -ENOENT; | |
1414 | ||
c8fe8f30 CH |
1415 | switch (inode->i_mode & S_IFMT) { |
1416 | case S_IFLNK: | |
1da177e4 | 1417 | return -ELOOP; |
c8fe8f30 CH |
1418 | case S_IFDIR: |
1419 | if (acc_mode & MAY_WRITE) | |
1420 | return -EISDIR; | |
1421 | break; | |
1422 | case S_IFBLK: | |
1423 | case S_IFCHR: | |
3fb64190 | 1424 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1425 | return -EACCES; |
c8fe8f30 CH |
1426 | /*FALLTHRU*/ |
1427 | case S_IFIFO: | |
1428 | case S_IFSOCK: | |
1da177e4 | 1429 | flag &= ~O_TRUNC; |
c8fe8f30 | 1430 | break; |
4a3fd211 | 1431 | } |
b41572e9 | 1432 | |
3fb64190 | 1433 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1434 | if (error) |
1435 | return error; | |
6146f0d5 | 1436 | |
1da177e4 LT |
1437 | /* |
1438 | * An append-only file must be opened in append mode for writing. | |
1439 | */ | |
1440 | if (IS_APPEND(inode)) { | |
8737c930 | 1441 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 1442 | return -EPERM; |
1da177e4 | 1443 | if (flag & O_TRUNC) |
7715b521 | 1444 | return -EPERM; |
1da177e4 LT |
1445 | } |
1446 | ||
1447 | /* O_NOATIME can only be set by the owner or superuser */ | |
7715b521 AV |
1448 | if (flag & O_NOATIME && !is_owner_or_cap(inode)) |
1449 | return -EPERM; | |
1da177e4 LT |
1450 | |
1451 | /* | |
1452 | * Ensure there are no outstanding leases on the file. | |
1453 | */ | |
b65a9cfc | 1454 | return break_lease(inode, flag); |
7715b521 | 1455 | } |
1da177e4 | 1456 | |
7715b521 AV |
1457 | static int handle_truncate(struct path *path) |
1458 | { | |
1459 | struct inode *inode = path->dentry->d_inode; | |
1460 | int error = get_write_access(inode); | |
1461 | if (error) | |
1462 | return error; | |
1463 | /* | |
1464 | * Refuse to truncate files with mandatory locks held on them. | |
1465 | */ | |
1466 | error = locks_verify_locked(inode); | |
1467 | if (!error) | |
ea0d3ab2 | 1468 | error = security_path_truncate(path); |
7715b521 AV |
1469 | if (!error) { |
1470 | error = do_truncate(path->dentry, 0, | |
1471 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1472 | NULL); | |
1473 | } | |
1474 | put_write_access(inode); | |
acd0c935 | 1475 | return error; |
1da177e4 LT |
1476 | } |
1477 | ||
d57999e1 DH |
1478 | /* |
1479 | * Be careful about ever adding any more callers of this | |
1480 | * function. Its flags must be in the namei format, not | |
1481 | * what get passed to sys_open(). | |
1482 | */ | |
1483 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
8737c930 | 1484 | int open_flag, int mode) |
aab520e2 DH |
1485 | { |
1486 | int error; | |
4ac91378 | 1487 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1488 | |
1489 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1490 | mode &= ~current_umask(); |
be6d3e56 KT |
1491 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1492 | if (error) | |
1493 | goto out_unlock; | |
aab520e2 | 1494 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1495 | out_unlock: |
aab520e2 | 1496 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1497 | dput(nd->path.dentry); |
1498 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1499 | if (error) |
1500 | return error; | |
1501 | /* Don't check for write permission, don't truncate */ | |
8737c930 | 1502 | return may_open(&nd->path, 0, open_flag & ~O_TRUNC); |
aab520e2 DH |
1503 | } |
1504 | ||
d57999e1 DH |
1505 | /* |
1506 | * Note that while the flag value (low two bits) for sys_open means: | |
1507 | * 00 - read-only | |
1508 | * 01 - write-only | |
1509 | * 10 - read-write | |
1510 | * 11 - special | |
1511 | * it is changed into | |
1512 | * 00 - no permissions needed | |
1513 | * 01 - read-permission | |
1514 | * 10 - write-permission | |
1515 | * 11 - read-write | |
1516 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1517 | * This is more logical, and also allows the 00 "no perm needed" | |
1518 | * to be used for symlinks (where the permissions are checked | |
1519 | * later). | |
1520 | * | |
1521 | */ | |
1522 | static inline int open_to_namei_flags(int flag) | |
1523 | { | |
1524 | if ((flag+1) & O_ACCMODE) | |
1525 | flag++; | |
1526 | return flag; | |
1527 | } | |
1528 | ||
7715b521 | 1529 | static int open_will_truncate(int flag, struct inode *inode) |
4a3fd211 DH |
1530 | { |
1531 | /* | |
1532 | * We'll never write to the fs underlying | |
1533 | * a device file. | |
1534 | */ | |
1535 | if (special_file(inode->i_mode)) | |
1536 | return 0; | |
1537 | return (flag & O_TRUNC); | |
1538 | } | |
1539 | ||
648fa861 | 1540 | static struct file *finish_open(struct nameidata *nd, |
9a66179e | 1541 | int open_flag, int acc_mode) |
648fa861 AV |
1542 | { |
1543 | struct file *filp; | |
1544 | int will_truncate; | |
1545 | int error; | |
1546 | ||
9a66179e | 1547 | will_truncate = open_will_truncate(open_flag, nd->path.dentry->d_inode); |
648fa861 AV |
1548 | if (will_truncate) { |
1549 | error = mnt_want_write(nd->path.mnt); | |
1550 | if (error) | |
1551 | goto exit; | |
1552 | } | |
1553 | error = may_open(&nd->path, acc_mode, open_flag); | |
1554 | if (error) { | |
1555 | if (will_truncate) | |
1556 | mnt_drop_write(nd->path.mnt); | |
1557 | goto exit; | |
1558 | } | |
1559 | filp = nameidata_to_filp(nd); | |
1560 | if (!IS_ERR(filp)) { | |
1561 | error = ima_file_check(filp, acc_mode); | |
1562 | if (error) { | |
1563 | fput(filp); | |
1564 | filp = ERR_PTR(error); | |
1565 | } | |
1566 | } | |
1567 | if (!IS_ERR(filp)) { | |
648fa861 AV |
1568 | if (will_truncate) { |
1569 | error = handle_truncate(&nd->path); | |
1570 | if (error) { | |
1571 | fput(filp); | |
1572 | filp = ERR_PTR(error); | |
1573 | } | |
1574 | } | |
1575 | } | |
1576 | /* | |
1577 | * It is now safe to drop the mnt write | |
1578 | * because the filp has had a write taken | |
1579 | * on its behalf. | |
1580 | */ | |
1581 | if (will_truncate) | |
1582 | mnt_drop_write(nd->path.mnt); | |
1583 | return filp; | |
1584 | ||
1585 | exit: | |
1586 | if (!IS_ERR(nd->intent.open.file)) | |
1587 | release_open_intent(nd); | |
1588 | path_put(&nd->path); | |
1589 | return ERR_PTR(error); | |
1590 | } | |
1591 | ||
fb1cc555 | 1592 | static struct file *do_last(struct nameidata *nd, struct path *path, |
5b369df8 | 1593 | int open_flag, int acc_mode, |
3e297b61 | 1594 | int mode, const char *pathname) |
fb1cc555 | 1595 | { |
a1e28038 | 1596 | struct dentry *dir = nd->path.dentry; |
fb1cc555 | 1597 | struct file *filp; |
1f36f774 AV |
1598 | int error = -EISDIR; |
1599 | ||
1600 | switch (nd->last_type) { | |
1601 | case LAST_DOTDOT: | |
1602 | follow_dotdot(nd); | |
1603 | dir = nd->path.dentry; | |
176306f5 | 1604 | case LAST_DOT: |
1f36f774 AV |
1605 | if (nd->path.mnt->mnt_sb->s_type->fs_flags & FS_REVAL_DOT) { |
1606 | if (!dir->d_op->d_revalidate(dir, nd)) { | |
1607 | error = -ESTALE; | |
1608 | goto exit; | |
1609 | } | |
1610 | } | |
1611 | /* fallthrough */ | |
1f36f774 AV |
1612 | case LAST_ROOT: |
1613 | if (open_flag & O_CREAT) | |
1614 | goto exit; | |
1615 | /* fallthrough */ | |
1616 | case LAST_BIND: | |
1617 | audit_inode(pathname, dir); | |
67ee3ad2 | 1618 | goto ok; |
1f36f774 | 1619 | } |
67ee3ad2 | 1620 | |
1f36f774 AV |
1621 | /* trailing slashes? */ |
1622 | if (nd->last.name[nd->last.len]) { | |
1623 | if (open_flag & O_CREAT) | |
1624 | goto exit; | |
002baeec | 1625 | nd->flags |= LOOKUP_DIRECTORY | LOOKUP_FOLLOW; |
1f36f774 AV |
1626 | } |
1627 | ||
1628 | /* just plain open? */ | |
1629 | if (!(open_flag & O_CREAT)) { | |
1630 | error = do_lookup(nd, &nd->last, path); | |
1631 | if (error) | |
1632 | goto exit; | |
1633 | error = -ENOENT; | |
1634 | if (!path->dentry->d_inode) | |
1635 | goto exit_dput; | |
1636 | if (path->dentry->d_inode->i_op->follow_link) | |
1637 | return NULL; | |
1638 | error = -ENOTDIR; | |
3e297b61 AV |
1639 | if (nd->flags & LOOKUP_DIRECTORY) { |
1640 | if (!path->dentry->d_inode->i_op->lookup) | |
1641 | goto exit_dput; | |
1642 | } | |
1f36f774 AV |
1643 | path_to_nameidata(path, nd); |
1644 | audit_inode(pathname, nd->path.dentry); | |
1645 | goto ok; | |
1646 | } | |
a2c36b45 | 1647 | |
1f36f774 | 1648 | /* OK, it's O_CREAT */ |
a1e28038 AV |
1649 | mutex_lock(&dir->d_inode->i_mutex); |
1650 | ||
1651 | path->dentry = lookup_hash(nd); | |
1652 | path->mnt = nd->path.mnt; | |
1653 | ||
fb1cc555 AV |
1654 | error = PTR_ERR(path->dentry); |
1655 | if (IS_ERR(path->dentry)) { | |
1656 | mutex_unlock(&dir->d_inode->i_mutex); | |
1657 | goto exit; | |
1658 | } | |
1659 | ||
1660 | if (IS_ERR(nd->intent.open.file)) { | |
1661 | error = PTR_ERR(nd->intent.open.file); | |
1662 | goto exit_mutex_unlock; | |
1663 | } | |
1664 | ||
1665 | /* Negative dentry, just create the file */ | |
1666 | if (!path->dentry->d_inode) { | |
1667 | /* | |
1668 | * This write is needed to ensure that a | |
1669 | * ro->rw transition does not occur between | |
1670 | * the time when the file is created and when | |
1671 | * a permanent write count is taken through | |
1672 | * the 'struct file' in nameidata_to_filp(). | |
1673 | */ | |
1674 | error = mnt_want_write(nd->path.mnt); | |
1675 | if (error) | |
1676 | goto exit_mutex_unlock; | |
1677 | error = __open_namei_create(nd, path, open_flag, mode); | |
1678 | if (error) { | |
1679 | mnt_drop_write(nd->path.mnt); | |
1680 | goto exit; | |
1681 | } | |
1682 | filp = nameidata_to_filp(nd); | |
1683 | mnt_drop_write(nd->path.mnt); | |
fb1cc555 AV |
1684 | if (!IS_ERR(filp)) { |
1685 | error = ima_file_check(filp, acc_mode); | |
1686 | if (error) { | |
1687 | fput(filp); | |
1688 | filp = ERR_PTR(error); | |
1689 | } | |
1690 | } | |
1691 | return filp; | |
1692 | } | |
1693 | ||
1694 | /* | |
1695 | * It already exists. | |
1696 | */ | |
1697 | mutex_unlock(&dir->d_inode->i_mutex); | |
1698 | audit_inode(pathname, path->dentry); | |
1699 | ||
1700 | error = -EEXIST; | |
5b369df8 | 1701 | if (open_flag & O_EXCL) |
fb1cc555 AV |
1702 | goto exit_dput; |
1703 | ||
1704 | if (__follow_mount(path)) { | |
1705 | error = -ELOOP; | |
5b369df8 | 1706 | if (open_flag & O_NOFOLLOW) |
fb1cc555 AV |
1707 | goto exit_dput; |
1708 | } | |
1709 | ||
1710 | error = -ENOENT; | |
1711 | if (!path->dentry->d_inode) | |
1712 | goto exit_dput; | |
9e67f361 AV |
1713 | |
1714 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 1715 | return NULL; |
fb1cc555 AV |
1716 | |
1717 | path_to_nameidata(path, nd); | |
1718 | error = -EISDIR; | |
1719 | if (S_ISDIR(path->dentry->d_inode->i_mode)) | |
1720 | goto exit; | |
67ee3ad2 | 1721 | ok: |
9a66179e | 1722 | filp = finish_open(nd, open_flag, acc_mode); |
fb1cc555 AV |
1723 | return filp; |
1724 | ||
1725 | exit_mutex_unlock: | |
1726 | mutex_unlock(&dir->d_inode->i_mutex); | |
1727 | exit_dput: | |
1728 | path_put_conditional(path, nd); | |
1729 | exit: | |
1730 | if (!IS_ERR(nd->intent.open.file)) | |
1731 | release_open_intent(nd); | |
fb1cc555 AV |
1732 | path_put(&nd->path); |
1733 | return ERR_PTR(error); | |
1734 | } | |
1735 | ||
1da177e4 | 1736 | /* |
4a3fd211 DH |
1737 | * Note that the low bits of the passed in "open_flag" |
1738 | * are not the same as in the local variable "flag". See | |
1739 | * open_to_namei_flags() for more details. | |
1da177e4 | 1740 | */ |
a70e65df | 1741 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1742 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1743 | { |
4a3fd211 | 1744 | struct file *filp; |
a70e65df | 1745 | struct nameidata nd; |
6e8341a1 | 1746 | int error; |
9850c056 | 1747 | struct path path; |
1da177e4 | 1748 | int count = 0; |
d57999e1 | 1749 | int flag = open_to_namei_flags(open_flag); |
9850c056 | 1750 | int force_reval = 0; |
1f36f774 AV |
1751 | |
1752 | if (!(open_flag & O_CREAT)) | |
1753 | mode = 0; | |
1da177e4 | 1754 | |
6b2f3d1f CH |
1755 | /* |
1756 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1757 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1758 | * always set instead of having to deal with possibly weird behaviour | |
1759 | * for malicious applications setting only __O_SYNC. | |
1760 | */ | |
1761 | if (open_flag & __O_SYNC) | |
1762 | open_flag |= O_DSYNC; | |
1763 | ||
6e8341a1 | 1764 | if (!acc_mode) |
6d125529 | 1765 | acc_mode = MAY_OPEN | ACC_MODE(open_flag); |
1da177e4 | 1766 | |
834f2a4a | 1767 | /* O_TRUNC implies we need access checks for write permissions */ |
4296e2cb | 1768 | if (open_flag & O_TRUNC) |
834f2a4a TM |
1769 | acc_mode |= MAY_WRITE; |
1770 | ||
1da177e4 LT |
1771 | /* Allow the LSM permission hook to distinguish append |
1772 | access from general write access. */ | |
4296e2cb | 1773 | if (open_flag & O_APPEND) |
1da177e4 LT |
1774 | acc_mode |= MAY_APPEND; |
1775 | ||
1f36f774 | 1776 | /* find the parent */ |
9850c056 | 1777 | reval: |
9b4a9b14 | 1778 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1779 | if (error) |
a70e65df | 1780 | return ERR_PTR(error); |
9850c056 AV |
1781 | if (force_reval) |
1782 | nd.flags |= LOOKUP_REVAL; | |
3866248e AV |
1783 | |
1784 | current->total_link_count = 0; | |
1785 | error = link_path_walk(pathname, &nd); | |
654f562c | 1786 | if (error) { |
10fa8e62 AV |
1787 | filp = ERR_PTR(error); |
1788 | goto out; | |
654f562c | 1789 | } |
1f36f774 | 1790 | if (unlikely(!audit_dummy_context()) && (open_flag & O_CREAT)) |
9b4a9b14 | 1791 | audit_inode(pathname, nd.path.dentry); |
1da177e4 LT |
1792 | |
1793 | /* | |
a2c36b45 | 1794 | * We have the parent and last component. |
1da177e4 | 1795 | */ |
1da177e4 | 1796 | |
8737f3a1 AV |
1797 | error = -ENFILE; |
1798 | filp = get_empty_filp(); | |
1799 | if (filp == NULL) | |
1800 | goto exit_parent; | |
1801 | nd.intent.open.file = filp; | |
482928d5 | 1802 | filp->f_flags = open_flag; |
8737f3a1 AV |
1803 | nd.intent.open.flags = flag; |
1804 | nd.intent.open.create_mode = mode; | |
a70e65df | 1805 | nd.flags &= ~LOOKUP_PARENT; |
1f36f774 AV |
1806 | nd.flags |= LOOKUP_OPEN; |
1807 | if (open_flag & O_CREAT) { | |
1808 | nd.flags |= LOOKUP_CREATE; | |
1809 | if (open_flag & O_EXCL) | |
1810 | nd.flags |= LOOKUP_EXCL; | |
1811 | } | |
3e297b61 AV |
1812 | if (open_flag & O_DIRECTORY) |
1813 | nd.flags |= LOOKUP_DIRECTORY; | |
002baeec JK |
1814 | if (!(open_flag & O_NOFOLLOW)) |
1815 | nd.flags |= LOOKUP_FOLLOW; | |
3e297b61 | 1816 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
806b681c | 1817 | while (unlikely(!filp)) { /* trailing symlink */ |
def4af30 | 1818 | struct path holder; |
1f36f774 | 1819 | struct inode *inode = path.dentry->d_inode; |
def4af30 | 1820 | void *cookie; |
806b681c | 1821 | error = -ELOOP; |
1f36f774 | 1822 | /* S_ISDIR part is a temporary automount kludge */ |
002baeec | 1823 | if (!(nd.flags & LOOKUP_FOLLOW) && !S_ISDIR(inode->i_mode)) |
1f36f774 AV |
1824 | goto exit_dput; |
1825 | if (count++ == 32) | |
806b681c AV |
1826 | goto exit_dput; |
1827 | /* | |
1828 | * This is subtle. Instead of calling do_follow_link() we do | |
1829 | * the thing by hands. The reason is that this way we have zero | |
1830 | * link_count and path_walk() (called from ->follow_link) | |
1831 | * honoring LOOKUP_PARENT. After that we have the parent and | |
1832 | * last component, i.e. we are in the same situation as after | |
1833 | * the first path_walk(). Well, almost - if the last component | |
1834 | * is normal we get its copy stored in nd->last.name and we will | |
1835 | * have to putname() it when we are done. Procfs-like symlinks | |
1836 | * just set LAST_BIND. | |
1837 | */ | |
1838 | nd.flags |= LOOKUP_PARENT; | |
1839 | error = security_inode_follow_link(path.dentry, &nd); | |
1840 | if (error) | |
1841 | goto exit_dput; | |
def4af30 AV |
1842 | error = __do_follow_link(&path, &nd, &cookie); |
1843 | if (unlikely(error)) { | |
806b681c | 1844 | /* nd.path had been dropped */ |
def4af30 AV |
1845 | if (!IS_ERR(cookie) && inode->i_op->put_link) |
1846 | inode->i_op->put_link(path.dentry, &nd, cookie); | |
1847 | path_put(&path); | |
806b681c AV |
1848 | release_open_intent(&nd); |
1849 | filp = ERR_PTR(error); | |
1850 | goto out; | |
1851 | } | |
def4af30 | 1852 | holder = path; |
806b681c | 1853 | nd.flags &= ~LOOKUP_PARENT; |
3e297b61 | 1854 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
def4af30 AV |
1855 | if (inode->i_op->put_link) |
1856 | inode->i_op->put_link(holder.dentry, &nd, cookie); | |
1857 | path_put(&holder); | |
806b681c | 1858 | } |
10fa8e62 | 1859 | out: |
2a737871 AV |
1860 | if (nd.root.mnt) |
1861 | path_put(&nd.root); | |
10fa8e62 AV |
1862 | if (filp == ERR_PTR(-ESTALE) && !force_reval) { |
1863 | force_reval = 1; | |
1864 | goto reval; | |
1865 | } | |
1866 | return filp; | |
1da177e4 | 1867 | |
806b681c AV |
1868 | exit_dput: |
1869 | path_put_conditional(&path, &nd); | |
1870 | if (!IS_ERR(nd.intent.open.file)) | |
a70e65df | 1871 | release_open_intent(&nd); |
806b681c AV |
1872 | exit_parent: |
1873 | path_put(&nd.path); | |
1874 | filp = ERR_PTR(error); | |
10fa8e62 | 1875 | goto out; |
1da177e4 LT |
1876 | } |
1877 | ||
a70e65df CH |
1878 | /** |
1879 | * filp_open - open file and return file pointer | |
1880 | * | |
1881 | * @filename: path to open | |
1882 | * @flags: open flags as per the open(2) second argument | |
1883 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1884 | * | |
1885 | * This is the helper to open a file from kernelspace if you really | |
1886 | * have to. But in generally you should not do this, so please move | |
1887 | * along, nothing to see here.. | |
1888 | */ | |
1889 | struct file *filp_open(const char *filename, int flags, int mode) | |
1890 | { | |
6e8341a1 | 1891 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1892 | } |
1893 | EXPORT_SYMBOL(filp_open); | |
1894 | ||
1da177e4 LT |
1895 | /** |
1896 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1897 | * @nd: nameidata info | |
1898 | * @is_dir: directory flag | |
1899 | * | |
1900 | * Simple function to lookup and return a dentry and create it | |
1901 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1902 | * |
4ac91378 | 1903 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1904 | */ |
1905 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1906 | { | |
c663e5d8 | 1907 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1908 | |
4ac91378 | 1909 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1910 | /* |
1911 | * Yucky last component or no last component at all? | |
1912 | * (foo/., foo/.., /////) | |
1913 | */ | |
1da177e4 LT |
1914 | if (nd->last_type != LAST_NORM) |
1915 | goto fail; | |
1916 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1917 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1918 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1919 | |
1920 | /* | |
1921 | * Do the final lookup. | |
1922 | */ | |
49705b77 | 1923 | dentry = lookup_hash(nd); |
1da177e4 LT |
1924 | if (IS_ERR(dentry)) |
1925 | goto fail; | |
c663e5d8 | 1926 | |
e9baf6e5 AV |
1927 | if (dentry->d_inode) |
1928 | goto eexist; | |
c663e5d8 CH |
1929 | /* |
1930 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1931 | * From the vfs_mknod() POV we just have a negative dentry - | |
1932 | * all is fine. Let's be bastards - you had / on the end, you've | |
1933 | * been asking for (non-existent) directory. -ENOENT for you. | |
1934 | */ | |
e9baf6e5 AV |
1935 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1936 | dput(dentry); | |
1937 | dentry = ERR_PTR(-ENOENT); | |
1938 | } | |
1da177e4 | 1939 | return dentry; |
e9baf6e5 | 1940 | eexist: |
1da177e4 | 1941 | dput(dentry); |
e9baf6e5 | 1942 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1943 | fail: |
1944 | return dentry; | |
1945 | } | |
f81a0bff | 1946 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1947 | |
1948 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1949 | { | |
a95164d9 | 1950 | int error = may_create(dir, dentry); |
1da177e4 LT |
1951 | |
1952 | if (error) | |
1953 | return error; | |
1954 | ||
1955 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1956 | return -EPERM; | |
1957 | ||
acfa4380 | 1958 | if (!dir->i_op->mknod) |
1da177e4 LT |
1959 | return -EPERM; |
1960 | ||
08ce5f16 SH |
1961 | error = devcgroup_inode_mknod(mode, dev); |
1962 | if (error) | |
1963 | return error; | |
1964 | ||
1da177e4 LT |
1965 | error = security_inode_mknod(dir, dentry, mode, dev); |
1966 | if (error) | |
1967 | return error; | |
1968 | ||
1da177e4 | 1969 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1970 | if (!error) |
f38aa942 | 1971 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1972 | return error; |
1973 | } | |
1974 | ||
463c3197 DH |
1975 | static int may_mknod(mode_t mode) |
1976 | { | |
1977 | switch (mode & S_IFMT) { | |
1978 | case S_IFREG: | |
1979 | case S_IFCHR: | |
1980 | case S_IFBLK: | |
1981 | case S_IFIFO: | |
1982 | case S_IFSOCK: | |
1983 | case 0: /* zero mode translates to S_IFREG */ | |
1984 | return 0; | |
1985 | case S_IFDIR: | |
1986 | return -EPERM; | |
1987 | default: | |
1988 | return -EINVAL; | |
1989 | } | |
1990 | } | |
1991 | ||
2e4d0924 HC |
1992 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
1993 | unsigned, dev) | |
1da177e4 | 1994 | { |
2ad94ae6 AV |
1995 | int error; |
1996 | char *tmp; | |
1997 | struct dentry *dentry; | |
1da177e4 LT |
1998 | struct nameidata nd; |
1999 | ||
2000 | if (S_ISDIR(mode)) | |
2001 | return -EPERM; | |
1da177e4 | 2002 | |
2ad94ae6 | 2003 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2004 | if (error) |
2ad94ae6 AV |
2005 | return error; |
2006 | ||
1da177e4 | 2007 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2008 | if (IS_ERR(dentry)) { |
2009 | error = PTR_ERR(dentry); | |
2010 | goto out_unlock; | |
2011 | } | |
4ac91378 | 2012 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2013 | mode &= ~current_umask(); |
463c3197 DH |
2014 | error = may_mknod(mode); |
2015 | if (error) | |
2016 | goto out_dput; | |
2017 | error = mnt_want_write(nd.path.mnt); | |
2018 | if (error) | |
2019 | goto out_dput; | |
be6d3e56 KT |
2020 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2021 | if (error) | |
2022 | goto out_drop_write; | |
463c3197 | 2023 | switch (mode & S_IFMT) { |
1da177e4 | 2024 | case 0: case S_IFREG: |
4ac91378 | 2025 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2026 | break; |
2027 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2028 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2029 | new_decode_dev(dev)); |
2030 | break; | |
2031 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2032 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2033 | break; |
1da177e4 | 2034 | } |
be6d3e56 | 2035 | out_drop_write: |
463c3197 DH |
2036 | mnt_drop_write(nd.path.mnt); |
2037 | out_dput: | |
2038 | dput(dentry); | |
2039 | out_unlock: | |
4ac91378 | 2040 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2041 | path_put(&nd.path); |
1da177e4 LT |
2042 | putname(tmp); |
2043 | ||
2044 | return error; | |
2045 | } | |
2046 | ||
3480b257 | 2047 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2048 | { |
2049 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2050 | } | |
2051 | ||
1da177e4 LT |
2052 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2053 | { | |
a95164d9 | 2054 | int error = may_create(dir, dentry); |
1da177e4 LT |
2055 | |
2056 | if (error) | |
2057 | return error; | |
2058 | ||
acfa4380 | 2059 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2060 | return -EPERM; |
2061 | ||
2062 | mode &= (S_IRWXUGO|S_ISVTX); | |
2063 | error = security_inode_mkdir(dir, dentry, mode); | |
2064 | if (error) | |
2065 | return error; | |
2066 | ||
1da177e4 | 2067 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2068 | if (!error) |
f38aa942 | 2069 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2070 | return error; |
2071 | } | |
2072 | ||
2e4d0924 | 2073 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2074 | { |
2075 | int error = 0; | |
2076 | char * tmp; | |
6902d925 DH |
2077 | struct dentry *dentry; |
2078 | struct nameidata nd; | |
1da177e4 | 2079 | |
2ad94ae6 AV |
2080 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2081 | if (error) | |
6902d925 | 2082 | goto out_err; |
1da177e4 | 2083 | |
6902d925 DH |
2084 | dentry = lookup_create(&nd, 1); |
2085 | error = PTR_ERR(dentry); | |
2086 | if (IS_ERR(dentry)) | |
2087 | goto out_unlock; | |
1da177e4 | 2088 | |
4ac91378 | 2089 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2090 | mode &= ~current_umask(); |
463c3197 DH |
2091 | error = mnt_want_write(nd.path.mnt); |
2092 | if (error) | |
2093 | goto out_dput; | |
be6d3e56 KT |
2094 | error = security_path_mkdir(&nd.path, dentry, mode); |
2095 | if (error) | |
2096 | goto out_drop_write; | |
4ac91378 | 2097 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2098 | out_drop_write: |
463c3197 DH |
2099 | mnt_drop_write(nd.path.mnt); |
2100 | out_dput: | |
6902d925 DH |
2101 | dput(dentry); |
2102 | out_unlock: | |
4ac91378 | 2103 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2104 | path_put(&nd.path); |
6902d925 DH |
2105 | putname(tmp); |
2106 | out_err: | |
1da177e4 LT |
2107 | return error; |
2108 | } | |
2109 | ||
3cdad428 | 2110 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2111 | { |
2112 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2113 | } | |
2114 | ||
1da177e4 LT |
2115 | /* |
2116 | * We try to drop the dentry early: we should have | |
2117 | * a usage count of 2 if we're the only user of this | |
2118 | * dentry, and if that is true (possibly after pruning | |
2119 | * the dcache), then we drop the dentry now. | |
2120 | * | |
2121 | * A low-level filesystem can, if it choses, legally | |
2122 | * do a | |
2123 | * | |
2124 | * if (!d_unhashed(dentry)) | |
2125 | * return -EBUSY; | |
2126 | * | |
2127 | * if it cannot handle the case of removing a directory | |
2128 | * that is still in use by something else.. | |
2129 | */ | |
2130 | void dentry_unhash(struct dentry *dentry) | |
2131 | { | |
2132 | dget(dentry); | |
dc168427 | 2133 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2134 | spin_lock(&dcache_lock); |
2135 | spin_lock(&dentry->d_lock); | |
2136 | if (atomic_read(&dentry->d_count) == 2) | |
2137 | __d_drop(dentry); | |
2138 | spin_unlock(&dentry->d_lock); | |
2139 | spin_unlock(&dcache_lock); | |
2140 | } | |
2141 | ||
2142 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2143 | { | |
2144 | int error = may_delete(dir, dentry, 1); | |
2145 | ||
2146 | if (error) | |
2147 | return error; | |
2148 | ||
acfa4380 | 2149 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2150 | return -EPERM; |
2151 | ||
1b1dcc1b | 2152 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2153 | dentry_unhash(dentry); |
2154 | if (d_mountpoint(dentry)) | |
2155 | error = -EBUSY; | |
2156 | else { | |
2157 | error = security_inode_rmdir(dir, dentry); | |
2158 | if (!error) { | |
2159 | error = dir->i_op->rmdir(dir, dentry); | |
d83c49f3 | 2160 | if (!error) { |
1da177e4 | 2161 | dentry->d_inode->i_flags |= S_DEAD; |
d83c49f3 AV |
2162 | dont_mount(dentry); |
2163 | } | |
1da177e4 LT |
2164 | } |
2165 | } | |
1b1dcc1b | 2166 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2167 | if (!error) { |
1da177e4 LT |
2168 | d_delete(dentry); |
2169 | } | |
2170 | dput(dentry); | |
2171 | ||
2172 | return error; | |
2173 | } | |
2174 | ||
5590ff0d | 2175 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2176 | { |
2177 | int error = 0; | |
2178 | char * name; | |
2179 | struct dentry *dentry; | |
2180 | struct nameidata nd; | |
2181 | ||
2ad94ae6 | 2182 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2183 | if (error) |
2ad94ae6 | 2184 | return error; |
1da177e4 LT |
2185 | |
2186 | switch(nd.last_type) { | |
0612d9fb OH |
2187 | case LAST_DOTDOT: |
2188 | error = -ENOTEMPTY; | |
2189 | goto exit1; | |
2190 | case LAST_DOT: | |
2191 | error = -EINVAL; | |
2192 | goto exit1; | |
2193 | case LAST_ROOT: | |
2194 | error = -EBUSY; | |
2195 | goto exit1; | |
1da177e4 | 2196 | } |
0612d9fb OH |
2197 | |
2198 | nd.flags &= ~LOOKUP_PARENT; | |
2199 | ||
4ac91378 | 2200 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2201 | dentry = lookup_hash(&nd); |
1da177e4 | 2202 | error = PTR_ERR(dentry); |
6902d925 DH |
2203 | if (IS_ERR(dentry)) |
2204 | goto exit2; | |
0622753b DH |
2205 | error = mnt_want_write(nd.path.mnt); |
2206 | if (error) | |
2207 | goto exit3; | |
be6d3e56 KT |
2208 | error = security_path_rmdir(&nd.path, dentry); |
2209 | if (error) | |
2210 | goto exit4; | |
4ac91378 | 2211 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2212 | exit4: |
0622753b DH |
2213 | mnt_drop_write(nd.path.mnt); |
2214 | exit3: | |
6902d925 DH |
2215 | dput(dentry); |
2216 | exit2: | |
4ac91378 | 2217 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2218 | exit1: |
1d957f9b | 2219 | path_put(&nd.path); |
1da177e4 LT |
2220 | putname(name); |
2221 | return error; | |
2222 | } | |
2223 | ||
3cdad428 | 2224 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2225 | { |
2226 | return do_rmdir(AT_FDCWD, pathname); | |
2227 | } | |
2228 | ||
1da177e4 LT |
2229 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2230 | { | |
2231 | int error = may_delete(dir, dentry, 0); | |
2232 | ||
2233 | if (error) | |
2234 | return error; | |
2235 | ||
acfa4380 | 2236 | if (!dir->i_op->unlink) |
1da177e4 LT |
2237 | return -EPERM; |
2238 | ||
1b1dcc1b | 2239 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2240 | if (d_mountpoint(dentry)) |
2241 | error = -EBUSY; | |
2242 | else { | |
2243 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2244 | if (!error) { |
1da177e4 | 2245 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2246 | if (!error) |
d83c49f3 | 2247 | dont_mount(dentry); |
bec1052e | 2248 | } |
1da177e4 | 2249 | } |
1b1dcc1b | 2250 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2251 | |
2252 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2253 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2254 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2255 | d_delete(dentry); |
1da177e4 | 2256 | } |
0eeca283 | 2257 | |
1da177e4 LT |
2258 | return error; |
2259 | } | |
2260 | ||
2261 | /* | |
2262 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2263 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2264 | * writeout happening, and we don't want to prevent access to the directory |
2265 | * while waiting on the I/O. | |
2266 | */ | |
5590ff0d | 2267 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2268 | { |
2ad94ae6 AV |
2269 | int error; |
2270 | char *name; | |
1da177e4 LT |
2271 | struct dentry *dentry; |
2272 | struct nameidata nd; | |
2273 | struct inode *inode = NULL; | |
2274 | ||
2ad94ae6 | 2275 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2276 | if (error) |
2ad94ae6 AV |
2277 | return error; |
2278 | ||
1da177e4 LT |
2279 | error = -EISDIR; |
2280 | if (nd.last_type != LAST_NORM) | |
2281 | goto exit1; | |
0612d9fb OH |
2282 | |
2283 | nd.flags &= ~LOOKUP_PARENT; | |
2284 | ||
4ac91378 | 2285 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2286 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2287 | error = PTR_ERR(dentry); |
2288 | if (!IS_ERR(dentry)) { | |
2289 | /* Why not before? Because we want correct error value */ | |
2290 | if (nd.last.name[nd.last.len]) | |
2291 | goto slashes; | |
2292 | inode = dentry->d_inode; | |
2293 | if (inode) | |
2294 | atomic_inc(&inode->i_count); | |
0622753b DH |
2295 | error = mnt_want_write(nd.path.mnt); |
2296 | if (error) | |
2297 | goto exit2; | |
be6d3e56 KT |
2298 | error = security_path_unlink(&nd.path, dentry); |
2299 | if (error) | |
2300 | goto exit3; | |
4ac91378 | 2301 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2302 | exit3: |
0622753b | 2303 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2304 | exit2: |
2305 | dput(dentry); | |
2306 | } | |
4ac91378 | 2307 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2308 | if (inode) |
2309 | iput(inode); /* truncate the inode here */ | |
2310 | exit1: | |
1d957f9b | 2311 | path_put(&nd.path); |
1da177e4 LT |
2312 | putname(name); |
2313 | return error; | |
2314 | ||
2315 | slashes: | |
2316 | error = !dentry->d_inode ? -ENOENT : | |
2317 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2318 | goto exit2; | |
2319 | } | |
2320 | ||
2e4d0924 | 2321 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2322 | { |
2323 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2324 | return -EINVAL; | |
2325 | ||
2326 | if (flag & AT_REMOVEDIR) | |
2327 | return do_rmdir(dfd, pathname); | |
2328 | ||
2329 | return do_unlinkat(dfd, pathname); | |
2330 | } | |
2331 | ||
3480b257 | 2332 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2333 | { |
2334 | return do_unlinkat(AT_FDCWD, pathname); | |
2335 | } | |
2336 | ||
db2e747b | 2337 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2338 | { |
a95164d9 | 2339 | int error = may_create(dir, dentry); |
1da177e4 LT |
2340 | |
2341 | if (error) | |
2342 | return error; | |
2343 | ||
acfa4380 | 2344 | if (!dir->i_op->symlink) |
1da177e4 LT |
2345 | return -EPERM; |
2346 | ||
2347 | error = security_inode_symlink(dir, dentry, oldname); | |
2348 | if (error) | |
2349 | return error; | |
2350 | ||
1da177e4 | 2351 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2352 | if (!error) |
f38aa942 | 2353 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2354 | return error; |
2355 | } | |
2356 | ||
2e4d0924 HC |
2357 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2358 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2359 | { |
2ad94ae6 AV |
2360 | int error; |
2361 | char *from; | |
2362 | char *to; | |
6902d925 DH |
2363 | struct dentry *dentry; |
2364 | struct nameidata nd; | |
1da177e4 LT |
2365 | |
2366 | from = getname(oldname); | |
2ad94ae6 | 2367 | if (IS_ERR(from)) |
1da177e4 | 2368 | return PTR_ERR(from); |
1da177e4 | 2369 | |
2ad94ae6 | 2370 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2371 | if (error) |
2ad94ae6 AV |
2372 | goto out_putname; |
2373 | ||
6902d925 DH |
2374 | dentry = lookup_create(&nd, 0); |
2375 | error = PTR_ERR(dentry); | |
2376 | if (IS_ERR(dentry)) | |
2377 | goto out_unlock; | |
2378 | ||
75c3f29d DH |
2379 | error = mnt_want_write(nd.path.mnt); |
2380 | if (error) | |
2381 | goto out_dput; | |
be6d3e56 KT |
2382 | error = security_path_symlink(&nd.path, dentry, from); |
2383 | if (error) | |
2384 | goto out_drop_write; | |
db2e747b | 2385 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2386 | out_drop_write: |
75c3f29d DH |
2387 | mnt_drop_write(nd.path.mnt); |
2388 | out_dput: | |
6902d925 DH |
2389 | dput(dentry); |
2390 | out_unlock: | |
4ac91378 | 2391 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2392 | path_put(&nd.path); |
6902d925 DH |
2393 | putname(to); |
2394 | out_putname: | |
1da177e4 LT |
2395 | putname(from); |
2396 | return error; | |
2397 | } | |
2398 | ||
3480b257 | 2399 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2400 | { |
2401 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2402 | } | |
2403 | ||
1da177e4 LT |
2404 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2405 | { | |
2406 | struct inode *inode = old_dentry->d_inode; | |
2407 | int error; | |
2408 | ||
2409 | if (!inode) | |
2410 | return -ENOENT; | |
2411 | ||
a95164d9 | 2412 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2413 | if (error) |
2414 | return error; | |
2415 | ||
2416 | if (dir->i_sb != inode->i_sb) | |
2417 | return -EXDEV; | |
2418 | ||
2419 | /* | |
2420 | * A link to an append-only or immutable file cannot be created. | |
2421 | */ | |
2422 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2423 | return -EPERM; | |
acfa4380 | 2424 | if (!dir->i_op->link) |
1da177e4 | 2425 | return -EPERM; |
7e79eedb | 2426 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2427 | return -EPERM; |
2428 | ||
2429 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2430 | if (error) | |
2431 | return error; | |
2432 | ||
7e79eedb | 2433 | mutex_lock(&inode->i_mutex); |
1da177e4 | 2434 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2435 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2436 | if (!error) |
7e79eedb | 2437 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2438 | return error; |
2439 | } | |
2440 | ||
2441 | /* | |
2442 | * Hardlinks are often used in delicate situations. We avoid | |
2443 | * security-related surprises by not following symlinks on the | |
2444 | * newname. --KAB | |
2445 | * | |
2446 | * We don't follow them on the oldname either to be compatible | |
2447 | * with linux 2.0, and to avoid hard-linking to directories | |
2448 | * and other special files. --ADM | |
2449 | */ | |
2e4d0924 HC |
2450 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2451 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2452 | { |
2453 | struct dentry *new_dentry; | |
2d8f3038 AV |
2454 | struct nameidata nd; |
2455 | struct path old_path; | |
1da177e4 | 2456 | int error; |
2ad94ae6 | 2457 | char *to; |
1da177e4 | 2458 | |
45c9b11a | 2459 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2460 | return -EINVAL; |
2461 | ||
2d8f3038 AV |
2462 | error = user_path_at(olddfd, oldname, |
2463 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2464 | &old_path); | |
1da177e4 | 2465 | if (error) |
2ad94ae6 AV |
2466 | return error; |
2467 | ||
2468 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2469 | if (error) |
2470 | goto out; | |
2471 | error = -EXDEV; | |
2d8f3038 | 2472 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2473 | goto out_release; |
2474 | new_dentry = lookup_create(&nd, 0); | |
2475 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2476 | if (IS_ERR(new_dentry)) |
2477 | goto out_unlock; | |
75c3f29d DH |
2478 | error = mnt_want_write(nd.path.mnt); |
2479 | if (error) | |
2480 | goto out_dput; | |
be6d3e56 KT |
2481 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2482 | if (error) | |
2483 | goto out_drop_write; | |
2d8f3038 | 2484 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2485 | out_drop_write: |
75c3f29d DH |
2486 | mnt_drop_write(nd.path.mnt); |
2487 | out_dput: | |
6902d925 DH |
2488 | dput(new_dentry); |
2489 | out_unlock: | |
4ac91378 | 2490 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2491 | out_release: |
1d957f9b | 2492 | path_put(&nd.path); |
2ad94ae6 | 2493 | putname(to); |
1da177e4 | 2494 | out: |
2d8f3038 | 2495 | path_put(&old_path); |
1da177e4 LT |
2496 | |
2497 | return error; | |
2498 | } | |
2499 | ||
3480b257 | 2500 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2501 | { |
c04030e1 | 2502 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2503 | } |
2504 | ||
1da177e4 LT |
2505 | /* |
2506 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2507 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2508 | * Problems: | |
2509 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2510 | * b) race potential - two innocent renames can create a loop together. | |
2511 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2512 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2513 | * story. |
2514 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2515 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2516 | * whether the target exists). Solution: try to be smart with locking |
2517 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2518 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2519 | * move will be locked. Thus we can rank directories by the tree |
2520 | * (ancestors first) and rank all non-directories after them. | |
2521 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2522 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2523 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2524 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2525 | * we'd better make sure that there's no link(2) for them. | |
2526 | * d) some filesystems don't support opened-but-unlinked directories, | |
2527 | * either because of layout or because they are not ready to deal with | |
2528 | * all cases correctly. The latter will be fixed (taking this sort of | |
2529 | * stuff into VFS), but the former is not going away. Solution: the same | |
2530 | * trick as in rmdir(). | |
2531 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2532 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2533 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 2534 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
2535 | * locking]. |
2536 | */ | |
75c96f85 AB |
2537 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2538 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2539 | { |
2540 | int error = 0; | |
2541 | struct inode *target; | |
2542 | ||
2543 | /* | |
2544 | * If we are going to change the parent - check write permissions, | |
2545 | * we'll need to flip '..'. | |
2546 | */ | |
2547 | if (new_dir != old_dir) { | |
f419a2e3 | 2548 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2549 | if (error) |
2550 | return error; | |
2551 | } | |
2552 | ||
2553 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2554 | if (error) | |
2555 | return error; | |
2556 | ||
2557 | target = new_dentry->d_inode; | |
d83c49f3 | 2558 | if (target) |
1b1dcc1b | 2559 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2560 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2561 | error = -EBUSY; | |
d83c49f3 AV |
2562 | else { |
2563 | if (target) | |
2564 | dentry_unhash(new_dentry); | |
1da177e4 | 2565 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
d83c49f3 | 2566 | } |
1da177e4 | 2567 | if (target) { |
d83c49f3 | 2568 | if (!error) { |
1da177e4 | 2569 | target->i_flags |= S_DEAD; |
d83c49f3 AV |
2570 | dont_mount(new_dentry); |
2571 | } | |
1b1dcc1b | 2572 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2573 | if (d_unhashed(new_dentry)) |
2574 | d_rehash(new_dentry); | |
2575 | dput(new_dentry); | |
2576 | } | |
e31e14ec | 2577 | if (!error) |
349457cc MF |
2578 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2579 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2580 | return error; |
2581 | } | |
2582 | ||
75c96f85 AB |
2583 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2584 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2585 | { |
2586 | struct inode *target; | |
2587 | int error; | |
2588 | ||
2589 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2590 | if (error) | |
2591 | return error; | |
2592 | ||
2593 | dget(new_dentry); | |
2594 | target = new_dentry->d_inode; | |
2595 | if (target) | |
1b1dcc1b | 2596 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2597 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2598 | error = -EBUSY; | |
2599 | else | |
2600 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2601 | if (!error) { | |
bec1052e | 2602 | if (target) |
d83c49f3 | 2603 | dont_mount(new_dentry); |
349457cc | 2604 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2605 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2606 | } |
2607 | if (target) | |
1b1dcc1b | 2608 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2609 | dput(new_dentry); |
2610 | return error; | |
2611 | } | |
2612 | ||
2613 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2614 | struct inode *new_dir, struct dentry *new_dentry) | |
2615 | { | |
2616 | int error; | |
2617 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 2618 | const unsigned char *old_name; |
1da177e4 LT |
2619 | |
2620 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2621 | return 0; | |
2622 | ||
2623 | error = may_delete(old_dir, old_dentry, is_dir); | |
2624 | if (error) | |
2625 | return error; | |
2626 | ||
2627 | if (!new_dentry->d_inode) | |
a95164d9 | 2628 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2629 | else |
2630 | error = may_delete(new_dir, new_dentry, is_dir); | |
2631 | if (error) | |
2632 | return error; | |
2633 | ||
acfa4380 | 2634 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2635 | return -EPERM; |
2636 | ||
0eeca283 RL |
2637 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2638 | ||
1da177e4 LT |
2639 | if (is_dir) |
2640 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2641 | else | |
2642 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
2643 | if (!error) |
2644 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 2645 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
2646 | fsnotify_oldname_free(old_name); |
2647 | ||
1da177e4 LT |
2648 | return error; |
2649 | } | |
2650 | ||
2e4d0924 HC |
2651 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2652 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2653 | { |
2ad94ae6 AV |
2654 | struct dentry *old_dir, *new_dir; |
2655 | struct dentry *old_dentry, *new_dentry; | |
2656 | struct dentry *trap; | |
1da177e4 | 2657 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2658 | char *from; |
2659 | char *to; | |
2660 | int error; | |
1da177e4 | 2661 | |
2ad94ae6 | 2662 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2663 | if (error) |
2664 | goto exit; | |
2665 | ||
2ad94ae6 | 2666 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2667 | if (error) |
2668 | goto exit1; | |
2669 | ||
2670 | error = -EXDEV; | |
4ac91378 | 2671 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2672 | goto exit2; |
2673 | ||
4ac91378 | 2674 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2675 | error = -EBUSY; |
2676 | if (oldnd.last_type != LAST_NORM) | |
2677 | goto exit2; | |
2678 | ||
4ac91378 | 2679 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2680 | if (newnd.last_type != LAST_NORM) |
2681 | goto exit2; | |
2682 | ||
0612d9fb OH |
2683 | oldnd.flags &= ~LOOKUP_PARENT; |
2684 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2685 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2686 | |
1da177e4 LT |
2687 | trap = lock_rename(new_dir, old_dir); |
2688 | ||
49705b77 | 2689 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2690 | error = PTR_ERR(old_dentry); |
2691 | if (IS_ERR(old_dentry)) | |
2692 | goto exit3; | |
2693 | /* source must exist */ | |
2694 | error = -ENOENT; | |
2695 | if (!old_dentry->d_inode) | |
2696 | goto exit4; | |
2697 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2698 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2699 | error = -ENOTDIR; | |
2700 | if (oldnd.last.name[oldnd.last.len]) | |
2701 | goto exit4; | |
2702 | if (newnd.last.name[newnd.last.len]) | |
2703 | goto exit4; | |
2704 | } | |
2705 | /* source should not be ancestor of target */ | |
2706 | error = -EINVAL; | |
2707 | if (old_dentry == trap) | |
2708 | goto exit4; | |
49705b77 | 2709 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2710 | error = PTR_ERR(new_dentry); |
2711 | if (IS_ERR(new_dentry)) | |
2712 | goto exit4; | |
2713 | /* target should not be an ancestor of source */ | |
2714 | error = -ENOTEMPTY; | |
2715 | if (new_dentry == trap) | |
2716 | goto exit5; | |
2717 | ||
9079b1eb DH |
2718 | error = mnt_want_write(oldnd.path.mnt); |
2719 | if (error) | |
2720 | goto exit5; | |
be6d3e56 KT |
2721 | error = security_path_rename(&oldnd.path, old_dentry, |
2722 | &newnd.path, new_dentry); | |
2723 | if (error) | |
2724 | goto exit6; | |
1da177e4 LT |
2725 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2726 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2727 | exit6: |
9079b1eb | 2728 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2729 | exit5: |
2730 | dput(new_dentry); | |
2731 | exit4: | |
2732 | dput(old_dentry); | |
2733 | exit3: | |
2734 | unlock_rename(new_dir, old_dir); | |
2735 | exit2: | |
1d957f9b | 2736 | path_put(&newnd.path); |
2ad94ae6 | 2737 | putname(to); |
1da177e4 | 2738 | exit1: |
1d957f9b | 2739 | path_put(&oldnd.path); |
1da177e4 | 2740 | putname(from); |
2ad94ae6 | 2741 | exit: |
1da177e4 LT |
2742 | return error; |
2743 | } | |
2744 | ||
a26eab24 | 2745 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2746 | { |
2747 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2748 | } | |
2749 | ||
1da177e4 LT |
2750 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2751 | { | |
2752 | int len; | |
2753 | ||
2754 | len = PTR_ERR(link); | |
2755 | if (IS_ERR(link)) | |
2756 | goto out; | |
2757 | ||
2758 | len = strlen(link); | |
2759 | if (len > (unsigned) buflen) | |
2760 | len = buflen; | |
2761 | if (copy_to_user(buffer, link, len)) | |
2762 | len = -EFAULT; | |
2763 | out: | |
2764 | return len; | |
2765 | } | |
2766 | ||
2767 | /* | |
2768 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2769 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2770 | * using) it for any given inode is up to filesystem. | |
2771 | */ | |
2772 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2773 | { | |
2774 | struct nameidata nd; | |
cc314eef | 2775 | void *cookie; |
694a1764 | 2776 | int res; |
cc314eef | 2777 | |
1da177e4 | 2778 | nd.depth = 0; |
cc314eef | 2779 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2780 | if (IS_ERR(cookie)) |
2781 | return PTR_ERR(cookie); | |
2782 | ||
2783 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2784 | if (dentry->d_inode->i_op->put_link) | |
2785 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2786 | return res; | |
1da177e4 LT |
2787 | } |
2788 | ||
2789 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2790 | { | |
2791 | return __vfs_follow_link(nd, link); | |
2792 | } | |
2793 | ||
2794 | /* get the link contents into pagecache */ | |
2795 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2796 | { | |
ebd09abb DG |
2797 | char *kaddr; |
2798 | struct page *page; | |
1da177e4 | 2799 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2800 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2801 | if (IS_ERR(page)) |
6fe6900e | 2802 | return (char*)page; |
1da177e4 | 2803 | *ppage = page; |
ebd09abb DG |
2804 | kaddr = kmap(page); |
2805 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2806 | return kaddr; | |
1da177e4 LT |
2807 | } |
2808 | ||
2809 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2810 | { | |
2811 | struct page *page = NULL; | |
2812 | char *s = page_getlink(dentry, &page); | |
2813 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2814 | if (page) { | |
2815 | kunmap(page); | |
2816 | page_cache_release(page); | |
2817 | } | |
2818 | return res; | |
2819 | } | |
2820 | ||
cc314eef | 2821 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2822 | { |
cc314eef | 2823 | struct page *page = NULL; |
1da177e4 | 2824 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2825 | return page; |
1da177e4 LT |
2826 | } |
2827 | ||
cc314eef | 2828 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2829 | { |
cc314eef LT |
2830 | struct page *page = cookie; |
2831 | ||
2832 | if (page) { | |
1da177e4 LT |
2833 | kunmap(page); |
2834 | page_cache_release(page); | |
1da177e4 LT |
2835 | } |
2836 | } | |
2837 | ||
54566b2c NP |
2838 | /* |
2839 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2840 | */ | |
2841 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2842 | { |
2843 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2844 | struct page *page; |
afddba49 | 2845 | void *fsdata; |
beb497ab | 2846 | int err; |
1da177e4 | 2847 | char *kaddr; |
54566b2c NP |
2848 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2849 | if (nofs) | |
2850 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2851 | |
7e53cac4 | 2852 | retry: |
afddba49 | 2853 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2854 | flags, &page, &fsdata); |
1da177e4 | 2855 | if (err) |
afddba49 NP |
2856 | goto fail; |
2857 | ||
1da177e4 LT |
2858 | kaddr = kmap_atomic(page, KM_USER0); |
2859 | memcpy(kaddr, symname, len-1); | |
2860 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2861 | |
2862 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2863 | page, fsdata); | |
1da177e4 LT |
2864 | if (err < 0) |
2865 | goto fail; | |
afddba49 NP |
2866 | if (err < len-1) |
2867 | goto retry; | |
2868 | ||
1da177e4 LT |
2869 | mark_inode_dirty(inode); |
2870 | return 0; | |
1da177e4 LT |
2871 | fail: |
2872 | return err; | |
2873 | } | |
2874 | ||
0adb25d2 KK |
2875 | int page_symlink(struct inode *inode, const char *symname, int len) |
2876 | { | |
2877 | return __page_symlink(inode, symname, len, | |
54566b2c | 2878 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2879 | } |
2880 | ||
92e1d5be | 2881 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2882 | .readlink = generic_readlink, |
2883 | .follow_link = page_follow_link_light, | |
2884 | .put_link = page_put_link, | |
2885 | }; | |
2886 | ||
2d8f3038 | 2887 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2888 | EXPORT_SYMBOL(follow_down); |
2889 | EXPORT_SYMBOL(follow_up); | |
2890 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2891 | EXPORT_SYMBOL(getname); | |
2892 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2893 | EXPORT_SYMBOL(lookup_one_len); |
2894 | EXPORT_SYMBOL(page_follow_link_light); | |
2895 | EXPORT_SYMBOL(page_put_link); | |
2896 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2897 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2898 | EXPORT_SYMBOL(page_symlink); |
2899 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2900 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2901 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2902 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2903 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2904 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2905 | EXPORT_SYMBOL(unlock_rename); |
2906 | EXPORT_SYMBOL(vfs_create); | |
2907 | EXPORT_SYMBOL(vfs_follow_link); | |
2908 | EXPORT_SYMBOL(vfs_link); | |
2909 | EXPORT_SYMBOL(vfs_mkdir); | |
2910 | EXPORT_SYMBOL(vfs_mknod); | |
2911 | EXPORT_SYMBOL(generic_permission); | |
2912 | EXPORT_SYMBOL(vfs_readlink); | |
2913 | EXPORT_SYMBOL(vfs_rename); | |
2914 | EXPORT_SYMBOL(vfs_rmdir); | |
2915 | EXPORT_SYMBOL(vfs_symlink); | |
2916 | EXPORT_SYMBOL(vfs_unlink); | |
2917 | EXPORT_SYMBOL(dentry_unhash); | |
2918 | EXPORT_SYMBOL(generic_readlink); |