]>
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> | |
22 | #include <linux/quotaops.h> | |
23 | #include <linux/pagemap.h> | |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
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> |
1da177e4 LT |
34 | #include <asm/uaccess.h> |
35 | ||
36 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
37 | ||
38 | /* [Feb-1997 T. Schoebel-Theuer] | |
39 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
40 | * were necessary because of omirr. The reason is that omirr needs | |
41 | * to know the _real_ pathname, not the user-supplied one, in case | |
42 | * of symlinks (and also when transname replacements occur). | |
43 | * | |
44 | * The new code replaces the old recursive symlink resolution with | |
45 | * an iterative one (in case of non-nested symlink chains). It does | |
46 | * this with calls to <fs>_follow_link(). | |
47 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
48 | * replaced with a single function lookup_dentry() that can handle all | |
49 | * the special cases of the former code. | |
50 | * | |
51 | * With the new dcache, the pathname is stored at each inode, at least as | |
52 | * long as the refcount of the inode is positive. As a side effect, the | |
53 | * size of the dcache depends on the inode cache and thus is dynamic. | |
54 | * | |
55 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
56 | * resolution to correspond with current state of the code. | |
57 | * | |
58 | * Note that the symlink resolution is not *completely* iterative. | |
59 | * There is still a significant amount of tail- and mid- recursion in | |
60 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
61 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
62 | * may return different results than <fs>_follow_link(). Many virtual | |
63 | * filesystems (including /proc) exhibit this behavior. | |
64 | */ | |
65 | ||
66 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
67 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
68 | * and the name already exists in form of a symlink, try to create the new | |
69 | * name indicated by the symlink. The old code always complained that the | |
70 | * name already exists, due to not following the symlink even if its target | |
71 | * is nonexistent. The new semantics affects also mknod() and link() when | |
72 | * the name is a symlink pointing to a non-existant name. | |
73 | * | |
74 | * I don't know which semantics is the right one, since I have no access | |
75 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
76 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
77 | * "old" one. Personally, I think the new semantics is much more logical. | |
78 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
79 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
80 | * and in the old Linux semantics. | |
81 | */ | |
82 | ||
83 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
84 | * semantics. See the comments in "open_namei" and "do_link" below. | |
85 | * | |
86 | * [10-Sep-98 Alan Modra] Another symlink change. | |
87 | */ | |
88 | ||
89 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
90 | * inside the path - always follow. | |
91 | * in the last component in creation/removal/renaming - never follow. | |
92 | * if LOOKUP_FOLLOW passed - follow. | |
93 | * if the pathname has trailing slashes - follow. | |
94 | * otherwise - don't follow. | |
95 | * (applied in that order). | |
96 | * | |
97 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
98 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
99 | * During the 2.4 we need to fix the userland stuff depending on it - | |
100 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
101 | * XEmacs seems to be relying on it... | |
102 | */ | |
103 | /* | |
104 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 105 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
106 | * any extra contention... |
107 | */ | |
108 | ||
a02f76c3 | 109 | static int __link_path_walk(const char *name, struct nameidata *nd); |
c4a7808f | 110 | |
1da177e4 LT |
111 | /* In order to reduce some races, while at the same time doing additional |
112 | * checking and hopefully speeding things up, we copy filenames to the | |
113 | * kernel data space before using them.. | |
114 | * | |
115 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
116 | * PATH_MAX includes the nul terminator --RR. | |
117 | */ | |
858119e1 | 118 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
119 | { |
120 | int retval; | |
121 | unsigned long len = PATH_MAX; | |
122 | ||
123 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
124 | if ((unsigned long) filename >= TASK_SIZE) | |
125 | return -EFAULT; | |
126 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
127 | len = TASK_SIZE - (unsigned long) filename; | |
128 | } | |
129 | ||
130 | retval = strncpy_from_user(page, filename, len); | |
131 | if (retval > 0) { | |
132 | if (retval < len) | |
133 | return 0; | |
134 | return -ENAMETOOLONG; | |
135 | } else if (!retval) | |
136 | retval = -ENOENT; | |
137 | return retval; | |
138 | } | |
139 | ||
140 | char * getname(const char __user * filename) | |
141 | { | |
142 | char *tmp, *result; | |
143 | ||
144 | result = ERR_PTR(-ENOMEM); | |
145 | tmp = __getname(); | |
146 | if (tmp) { | |
147 | int retval = do_getname(filename, tmp); | |
148 | ||
149 | result = tmp; | |
150 | if (retval < 0) { | |
151 | __putname(tmp); | |
152 | result = ERR_PTR(retval); | |
153 | } | |
154 | } | |
155 | audit_getname(result); | |
156 | return result; | |
157 | } | |
158 | ||
159 | #ifdef CONFIG_AUDITSYSCALL | |
160 | void putname(const char *name) | |
161 | { | |
5ac3a9c2 | 162 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
163 | audit_putname(name); |
164 | else | |
165 | __putname(name); | |
166 | } | |
167 | EXPORT_SYMBOL(putname); | |
168 | #endif | |
169 | ||
170 | ||
171 | /** | |
172 | * generic_permission - check for access rights on a Posix-like filesystem | |
173 | * @inode: inode to check access rights for | |
174 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
175 | * @check_acl: optional callback to check for Posix ACLs | |
176 | * | |
177 | * Used to check for read/write/execute permissions on a file. | |
178 | * We use "fsuid" for this, letting us set arbitrary permissions | |
179 | * for filesystem access without changing the "normal" uids which | |
180 | * are used for other things.. | |
181 | */ | |
182 | int generic_permission(struct inode *inode, int mask, | |
183 | int (*check_acl)(struct inode *inode, int mask)) | |
184 | { | |
185 | umode_t mode = inode->i_mode; | |
186 | ||
e6305c43 AV |
187 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
188 | ||
1da177e4 LT |
189 | if (current->fsuid == inode->i_uid) |
190 | mode >>= 6; | |
191 | else { | |
192 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
193 | int error = check_acl(inode, mask); | |
194 | if (error == -EACCES) | |
195 | goto check_capabilities; | |
196 | else if (error != -EAGAIN) | |
197 | return error; | |
198 | } | |
199 | ||
200 | if (in_group_p(inode->i_gid)) | |
201 | mode >>= 3; | |
202 | } | |
203 | ||
204 | /* | |
205 | * If the DACs are ok we don't need any capability check. | |
206 | */ | |
e6305c43 | 207 | if ((mask & ~mode) == 0) |
1da177e4 LT |
208 | return 0; |
209 | ||
210 | check_capabilities: | |
211 | /* | |
212 | * Read/write DACs are always overridable. | |
213 | * Executable DACs are overridable if at least one exec bit is set. | |
214 | */ | |
215 | if (!(mask & MAY_EXEC) || | |
216 | (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode)) | |
217 | if (capable(CAP_DAC_OVERRIDE)) | |
218 | return 0; | |
219 | ||
220 | /* | |
221 | * Searching includes executable on directories, else just read. | |
222 | */ | |
223 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
224 | if (capable(CAP_DAC_READ_SEARCH)) | |
225 | return 0; | |
226 | ||
227 | return -EACCES; | |
228 | } | |
229 | ||
f419a2e3 | 230 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 231 | { |
e6305c43 | 232 | int retval; |
1da177e4 LT |
233 | |
234 | if (mask & MAY_WRITE) { | |
22590e41 | 235 | umode_t mode = inode->i_mode; |
1da177e4 LT |
236 | |
237 | /* | |
238 | * Nobody gets write access to a read-only fs. | |
239 | */ | |
240 | if (IS_RDONLY(inode) && | |
241 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
242 | return -EROFS; | |
243 | ||
244 | /* | |
245 | * Nobody gets write access to an immutable file. | |
246 | */ | |
247 | if (IS_IMMUTABLE(inode)) | |
248 | return -EACCES; | |
249 | } | |
250 | ||
1da177e4 | 251 | /* Ordinary permission routines do not understand MAY_APPEND. */ |
22590e41 | 252 | if (inode->i_op && inode->i_op->permission) { |
b77b0646 | 253 | retval = inode->i_op->permission(inode, mask); |
22590e41 MS |
254 | if (!retval) { |
255 | /* | |
256 | * Exec permission on a regular file is denied if none | |
257 | * of the execute bits are set. | |
258 | * | |
259 | * This check should be done by the ->permission() | |
260 | * method. | |
261 | */ | |
262 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode) && | |
263 | !(inode->i_mode & S_IXUGO)) | |
264 | return -EACCES; | |
265 | } | |
266 | } else { | |
e6305c43 | 267 | retval = generic_permission(inode, mask, NULL); |
22590e41 | 268 | } |
1da177e4 LT |
269 | if (retval) |
270 | return retval; | |
271 | ||
08ce5f16 SH |
272 | retval = devcgroup_inode_permission(inode, mask); |
273 | if (retval) | |
274 | return retval; | |
275 | ||
e6305c43 | 276 | return security_inode_permission(inode, |
f418b006 | 277 | mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND)); |
1da177e4 LT |
278 | } |
279 | ||
e4543edd CH |
280 | /** |
281 | * vfs_permission - check for access rights to a given path | |
282 | * @nd: lookup result that describes the path | |
283 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
284 | * | |
285 | * Used to check for read/write/execute permissions on a path. | |
286 | * We use "fsuid" for this, letting us set arbitrary permissions | |
287 | * for filesystem access without changing the "normal" uids which | |
288 | * are used for other things. | |
289 | */ | |
290 | int vfs_permission(struct nameidata *nd, int mask) | |
291 | { | |
f419a2e3 | 292 | return inode_permission(nd->path.dentry->d_inode, mask); |
e4543edd CH |
293 | } |
294 | ||
8c744fb8 CH |
295 | /** |
296 | * file_permission - check for additional access rights to a given file | |
297 | * @file: file to check access rights for | |
298 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
299 | * | |
300 | * Used to check for read/write/execute permissions on an already opened | |
301 | * file. | |
302 | * | |
303 | * Note: | |
304 | * Do not use this function in new code. All access checks should | |
305 | * be done using vfs_permission(). | |
306 | */ | |
307 | int file_permission(struct file *file, int mask) | |
308 | { | |
f419a2e3 | 309 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
310 | } |
311 | ||
1da177e4 LT |
312 | /* |
313 | * get_write_access() gets write permission for a file. | |
314 | * put_write_access() releases this write permission. | |
315 | * This is used for regular files. | |
316 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
317 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
318 | * can have the following values: | |
319 | * 0: no writers, no VM_DENYWRITE mappings | |
320 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
321 | * > 0: (i_writecount) users are writing to the file. | |
322 | * | |
323 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
324 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
325 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
326 | * to do the change if sign is wrong. Exclusion between them is provided by | |
327 | * the inode->i_lock spinlock. | |
328 | */ | |
329 | ||
330 | int get_write_access(struct inode * inode) | |
331 | { | |
332 | spin_lock(&inode->i_lock); | |
333 | if (atomic_read(&inode->i_writecount) < 0) { | |
334 | spin_unlock(&inode->i_lock); | |
335 | return -ETXTBSY; | |
336 | } | |
337 | atomic_inc(&inode->i_writecount); | |
338 | spin_unlock(&inode->i_lock); | |
339 | ||
340 | return 0; | |
341 | } | |
342 | ||
343 | int deny_write_access(struct file * file) | |
344 | { | |
0f7fc9e4 | 345 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
346 | |
347 | spin_lock(&inode->i_lock); | |
348 | if (atomic_read(&inode->i_writecount) > 0) { | |
349 | spin_unlock(&inode->i_lock); | |
350 | return -ETXTBSY; | |
351 | } | |
352 | atomic_dec(&inode->i_writecount); | |
353 | spin_unlock(&inode->i_lock); | |
354 | ||
355 | return 0; | |
356 | } | |
357 | ||
5dd784d0 JB |
358 | /** |
359 | * path_get - get a reference to a path | |
360 | * @path: path to get the reference to | |
361 | * | |
362 | * Given a path increment the reference count to the dentry and the vfsmount. | |
363 | */ | |
364 | void path_get(struct path *path) | |
365 | { | |
366 | mntget(path->mnt); | |
367 | dget(path->dentry); | |
368 | } | |
369 | EXPORT_SYMBOL(path_get); | |
370 | ||
1d957f9b JB |
371 | /** |
372 | * path_put - put a reference to a path | |
373 | * @path: path to put the reference to | |
374 | * | |
375 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
376 | */ | |
377 | void path_put(struct path *path) | |
1da177e4 | 378 | { |
1d957f9b JB |
379 | dput(path->dentry); |
380 | mntput(path->mnt); | |
1da177e4 | 381 | } |
1d957f9b | 382 | EXPORT_SYMBOL(path_put); |
1da177e4 | 383 | |
834f2a4a TM |
384 | /** |
385 | * release_open_intent - free up open intent resources | |
386 | * @nd: pointer to nameidata | |
387 | */ | |
388 | void release_open_intent(struct nameidata *nd) | |
389 | { | |
0f7fc9e4 | 390 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
391 | put_filp(nd->intent.open.file); |
392 | else | |
393 | fput(nd->intent.open.file); | |
394 | } | |
395 | ||
bcdc5e01 IK |
396 | static inline struct dentry * |
397 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
398 | { | |
399 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
400 | if (unlikely(status <= 0)) { | |
401 | /* | |
402 | * The dentry failed validation. | |
403 | * If d_revalidate returned 0 attempt to invalidate | |
404 | * the dentry otherwise d_revalidate is asking us | |
405 | * to return a fail status. | |
406 | */ | |
407 | if (!status) { | |
408 | if (!d_invalidate(dentry)) { | |
409 | dput(dentry); | |
410 | dentry = NULL; | |
411 | } | |
412 | } else { | |
413 | dput(dentry); | |
414 | dentry = ERR_PTR(status); | |
415 | } | |
416 | } | |
417 | return dentry; | |
418 | } | |
419 | ||
1da177e4 LT |
420 | /* |
421 | * Internal lookup() using the new generic dcache. | |
422 | * SMP-safe | |
423 | */ | |
424 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
425 | { | |
426 | struct dentry * dentry = __d_lookup(parent, name); | |
427 | ||
428 | /* lockess __d_lookup may fail due to concurrent d_move() | |
429 | * in some unrelated directory, so try with d_lookup | |
430 | */ | |
431 | if (!dentry) | |
432 | dentry = d_lookup(parent, name); | |
433 | ||
bcdc5e01 IK |
434 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
435 | dentry = do_revalidate(dentry, nd); | |
436 | ||
1da177e4 LT |
437 | return dentry; |
438 | } | |
439 | ||
440 | /* | |
441 | * Short-cut version of permission(), for calling by | |
442 | * path_walk(), when dcache lock is held. Combines parts | |
443 | * of permission() and generic_permission(), and tests ONLY for | |
444 | * MAY_EXEC permission. | |
445 | * | |
446 | * If appropriate, check DAC only. If not appropriate, or | |
447 | * short-cut DAC fails, then call permission() to do more | |
448 | * complete permission check. | |
449 | */ | |
672b16b2 | 450 | static int exec_permission_lite(struct inode *inode) |
1da177e4 LT |
451 | { |
452 | umode_t mode = inode->i_mode; | |
453 | ||
454 | if (inode->i_op && inode->i_op->permission) | |
455 | return -EAGAIN; | |
456 | ||
457 | if (current->fsuid == inode->i_uid) | |
458 | mode >>= 6; | |
459 | else if (in_group_p(inode->i_gid)) | |
460 | mode >>= 3; | |
461 | ||
462 | if (mode & MAY_EXEC) | |
463 | goto ok; | |
464 | ||
465 | if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE)) | |
466 | goto ok; | |
467 | ||
468 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE)) | |
469 | goto ok; | |
470 | ||
471 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH)) | |
472 | goto ok; | |
473 | ||
474 | return -EACCES; | |
475 | ok: | |
b77b0646 | 476 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
477 | } |
478 | ||
479 | /* | |
480 | * This is called when everything else fails, and we actually have | |
481 | * to go to the low-level filesystem to find out what we should do.. | |
482 | * | |
483 | * We get the directory semaphore, and after getting that we also | |
484 | * make sure that nobody added the entry to the dcache in the meantime.. | |
485 | * SMP-safe | |
486 | */ | |
487 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
488 | { | |
489 | struct dentry * result; | |
490 | struct inode *dir = parent->d_inode; | |
491 | ||
1b1dcc1b | 492 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
493 | /* |
494 | * First re-do the cached lookup just in case it was created | |
495 | * while we waited for the directory semaphore.. | |
496 | * | |
497 | * FIXME! This could use version numbering or similar to | |
498 | * avoid unnecessary cache lookups. | |
499 | * | |
500 | * The "dcache_lock" is purely to protect the RCU list walker | |
501 | * from concurrent renames at this point (we mustn't get false | |
502 | * negatives from the RCU list walk here, unlike the optimistic | |
503 | * fast walk). | |
504 | * | |
505 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
506 | */ | |
507 | result = d_lookup(parent, name); | |
508 | if (!result) { | |
d70b67c8 MS |
509 | struct dentry *dentry; |
510 | ||
511 | /* Don't create child dentry for a dead directory. */ | |
512 | result = ERR_PTR(-ENOENT); | |
513 | if (IS_DEADDIR(dir)) | |
514 | goto out_unlock; | |
515 | ||
516 | dentry = d_alloc(parent, name); | |
1da177e4 LT |
517 | result = ERR_PTR(-ENOMEM); |
518 | if (dentry) { | |
519 | result = dir->i_op->lookup(dir, dentry, nd); | |
520 | if (result) | |
521 | dput(dentry); | |
522 | else | |
523 | result = dentry; | |
524 | } | |
d70b67c8 | 525 | out_unlock: |
1b1dcc1b | 526 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
527 | return result; |
528 | } | |
529 | ||
530 | /* | |
531 | * Uhhuh! Nasty case: the cache was re-populated while | |
532 | * we waited on the semaphore. Need to revalidate. | |
533 | */ | |
1b1dcc1b | 534 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 535 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
536 | result = do_revalidate(result, nd); |
537 | if (!result) | |
1da177e4 | 538 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
539 | } |
540 | return result; | |
541 | } | |
542 | ||
1da177e4 | 543 | /* SMP-safe */ |
7f2da1e7 | 544 | static __always_inline void |
1da177e4 LT |
545 | walk_init_root(const char *name, struct nameidata *nd) |
546 | { | |
e518ddb7 AM |
547 | struct fs_struct *fs = current->fs; |
548 | ||
549 | read_lock(&fs->lock); | |
6ac08c39 JB |
550 | nd->path = fs->root; |
551 | path_get(&fs->root); | |
e518ddb7 | 552 | read_unlock(&fs->lock); |
1da177e4 LT |
553 | } |
554 | ||
a02f76c3 AV |
555 | /* |
556 | * Wrapper to retry pathname resolution whenever the underlying | |
557 | * file system returns an ESTALE. | |
558 | * | |
559 | * Retry the whole path once, forcing real lookup requests | |
560 | * instead of relying on the dcache. | |
561 | */ | |
562 | static __always_inline int link_path_walk(const char *name, struct nameidata *nd) | |
563 | { | |
564 | struct path save = nd->path; | |
565 | int result; | |
566 | ||
567 | /* make sure the stuff we saved doesn't go away */ | |
c8e7f449 | 568 | path_get(&save); |
a02f76c3 AV |
569 | |
570 | result = __link_path_walk(name, nd); | |
571 | if (result == -ESTALE) { | |
572 | /* nd->path had been dropped */ | |
573 | nd->path = save; | |
c8e7f449 | 574 | path_get(&nd->path); |
a02f76c3 AV |
575 | nd->flags |= LOOKUP_REVAL; |
576 | result = __link_path_walk(name, nd); | |
577 | } | |
578 | ||
579 | path_put(&save); | |
580 | ||
581 | return result; | |
582 | } | |
583 | ||
f1662356 | 584 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
585 | { |
586 | int res = 0; | |
587 | char *name; | |
588 | if (IS_ERR(link)) | |
589 | goto fail; | |
590 | ||
591 | if (*link == '/') { | |
1d957f9b | 592 | path_put(&nd->path); |
7f2da1e7 | 593 | walk_init_root(link, nd); |
1da177e4 LT |
594 | } |
595 | res = link_path_walk(link, nd); | |
1da177e4 LT |
596 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
597 | return res; | |
598 | /* | |
599 | * If it is an iterative symlinks resolution in open_namei() we | |
600 | * have to copy the last component. And all that crap because of | |
601 | * bloody create() on broken symlinks. Furrfu... | |
602 | */ | |
603 | name = __getname(); | |
604 | if (unlikely(!name)) { | |
1d957f9b | 605 | path_put(&nd->path); |
1da177e4 LT |
606 | return -ENOMEM; |
607 | } | |
608 | strcpy(name, nd->last.name); | |
609 | nd->last.name = name; | |
610 | return 0; | |
611 | fail: | |
1d957f9b | 612 | path_put(&nd->path); |
1da177e4 LT |
613 | return PTR_ERR(link); |
614 | } | |
615 | ||
1d957f9b | 616 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
617 | { |
618 | dput(path->dentry); | |
4ac91378 | 619 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
620 | mntput(path->mnt); |
621 | } | |
622 | ||
623 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
624 | { | |
4ac91378 JB |
625 | dput(nd->path.dentry); |
626 | if (nd->path.mnt != path->mnt) | |
627 | mntput(nd->path.mnt); | |
628 | nd->path.mnt = path->mnt; | |
629 | nd->path.dentry = path->dentry; | |
051d3812 IK |
630 | } |
631 | ||
f1662356 | 632 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
633 | { |
634 | int error; | |
cc314eef | 635 | void *cookie; |
cd4e91d3 | 636 | struct dentry *dentry = path->dentry; |
1da177e4 | 637 | |
d671a1cb | 638 | touch_atime(path->mnt, dentry); |
1da177e4 | 639 | nd_set_link(nd, NULL); |
cd4e91d3 | 640 | |
4ac91378 | 641 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
642 | path_to_nameidata(path, nd); |
643 | dget(dentry); | |
644 | } | |
645 | mntget(path->mnt); | |
cc314eef LT |
646 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
647 | error = PTR_ERR(cookie); | |
648 | if (!IS_ERR(cookie)) { | |
1da177e4 | 649 | char *s = nd_get_link(nd); |
cc314eef | 650 | error = 0; |
1da177e4 LT |
651 | if (s) |
652 | error = __vfs_follow_link(nd, s); | |
653 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 654 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 655 | } |
09da5916 | 656 | path_put(path); |
1da177e4 LT |
657 | |
658 | return error; | |
659 | } | |
660 | ||
661 | /* | |
662 | * This limits recursive symlink follows to 8, while | |
663 | * limiting consecutive symlinks to 40. | |
664 | * | |
665 | * Without that kind of total limit, nasty chains of consecutive | |
666 | * symlinks can cause almost arbitrarily long lookups. | |
667 | */ | |
90ebe565 | 668 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
669 | { |
670 | int err = -ELOOP; | |
671 | if (current->link_count >= MAX_NESTED_LINKS) | |
672 | goto loop; | |
673 | if (current->total_link_count >= 40) | |
674 | goto loop; | |
675 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
676 | cond_resched(); | |
90ebe565 | 677 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
678 | if (err) |
679 | goto loop; | |
680 | current->link_count++; | |
681 | current->total_link_count++; | |
682 | nd->depth++; | |
cd4e91d3 | 683 | err = __do_follow_link(path, nd); |
839d9f93 AV |
684 | current->link_count--; |
685 | nd->depth--; | |
1da177e4 LT |
686 | return err; |
687 | loop: | |
1d957f9b JB |
688 | path_put_conditional(path, nd); |
689 | path_put(&nd->path); | |
1da177e4 LT |
690 | return err; |
691 | } | |
692 | ||
693 | int follow_up(struct vfsmount **mnt, struct dentry **dentry) | |
694 | { | |
695 | struct vfsmount *parent; | |
696 | struct dentry *mountpoint; | |
697 | spin_lock(&vfsmount_lock); | |
698 | parent=(*mnt)->mnt_parent; | |
699 | if (parent == *mnt) { | |
700 | spin_unlock(&vfsmount_lock); | |
701 | return 0; | |
702 | } | |
703 | mntget(parent); | |
704 | mountpoint=dget((*mnt)->mnt_mountpoint); | |
705 | spin_unlock(&vfsmount_lock); | |
706 | dput(*dentry); | |
707 | *dentry = mountpoint; | |
708 | mntput(*mnt); | |
709 | *mnt = parent; | |
710 | return 1; | |
711 | } | |
712 | ||
713 | /* no need for dcache_lock, as serialization is taken care in | |
714 | * namespace.c | |
715 | */ | |
463ffb2e AV |
716 | static int __follow_mount(struct path *path) |
717 | { | |
718 | int res = 0; | |
719 | while (d_mountpoint(path->dentry)) { | |
720 | struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry); | |
721 | if (!mounted) | |
722 | break; | |
723 | dput(path->dentry); | |
724 | if (res) | |
725 | mntput(path->mnt); | |
726 | path->mnt = mounted; | |
727 | path->dentry = dget(mounted->mnt_root); | |
728 | res = 1; | |
729 | } | |
730 | return res; | |
731 | } | |
732 | ||
58c465eb | 733 | static void follow_mount(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 | 734 | { |
1da177e4 LT |
735 | while (d_mountpoint(*dentry)) { |
736 | struct vfsmount *mounted = lookup_mnt(*mnt, *dentry); | |
737 | if (!mounted) | |
738 | break; | |
58c465eb | 739 | dput(*dentry); |
1da177e4 LT |
740 | mntput(*mnt); |
741 | *mnt = mounted; | |
1da177e4 | 742 | *dentry = dget(mounted->mnt_root); |
1da177e4 | 743 | } |
1da177e4 LT |
744 | } |
745 | ||
746 | /* no need for dcache_lock, as serialization is taken care in | |
747 | * namespace.c | |
748 | */ | |
e13b210f | 749 | int follow_down(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 LT |
750 | { |
751 | struct vfsmount *mounted; | |
752 | ||
753 | mounted = lookup_mnt(*mnt, *dentry); | |
754 | if (mounted) { | |
e13b210f | 755 | dput(*dentry); |
1da177e4 LT |
756 | mntput(*mnt); |
757 | *mnt = mounted; | |
1da177e4 LT |
758 | *dentry = dget(mounted->mnt_root); |
759 | return 1; | |
760 | } | |
761 | return 0; | |
762 | } | |
763 | ||
f1662356 | 764 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 765 | { |
e518ddb7 AM |
766 | struct fs_struct *fs = current->fs; |
767 | ||
1da177e4 LT |
768 | while(1) { |
769 | struct vfsmount *parent; | |
4ac91378 | 770 | struct dentry *old = nd->path.dentry; |
1da177e4 | 771 | |
e518ddb7 | 772 | read_lock(&fs->lock); |
6ac08c39 JB |
773 | if (nd->path.dentry == fs->root.dentry && |
774 | nd->path.mnt == fs->root.mnt) { | |
e518ddb7 | 775 | read_unlock(&fs->lock); |
1da177e4 LT |
776 | break; |
777 | } | |
e518ddb7 | 778 | read_unlock(&fs->lock); |
1da177e4 | 779 | spin_lock(&dcache_lock); |
4ac91378 JB |
780 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
781 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
782 | spin_unlock(&dcache_lock); |
783 | dput(old); | |
784 | break; | |
785 | } | |
786 | spin_unlock(&dcache_lock); | |
787 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
788 | parent = nd->path.mnt->mnt_parent; |
789 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
790 | spin_unlock(&vfsmount_lock); |
791 | break; | |
792 | } | |
793 | mntget(parent); | |
4ac91378 | 794 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
795 | spin_unlock(&vfsmount_lock); |
796 | dput(old); | |
4ac91378 JB |
797 | mntput(nd->path.mnt); |
798 | nd->path.mnt = parent; | |
1da177e4 | 799 | } |
4ac91378 | 800 | follow_mount(&nd->path.mnt, &nd->path.dentry); |
1da177e4 LT |
801 | } |
802 | ||
1da177e4 LT |
803 | /* |
804 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
805 | * small and for now I'd prefer to have fast path as straight as possible. | |
806 | * It _is_ time-critical. | |
807 | */ | |
808 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
809 | struct path *path) | |
810 | { | |
4ac91378 JB |
811 | struct vfsmount *mnt = nd->path.mnt; |
812 | struct dentry *dentry = __d_lookup(nd->path.dentry, name); | |
1da177e4 LT |
813 | |
814 | if (!dentry) | |
815 | goto need_lookup; | |
816 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
817 | goto need_revalidate; | |
818 | done: | |
819 | path->mnt = mnt; | |
820 | path->dentry = dentry; | |
634ee701 | 821 | __follow_mount(path); |
1da177e4 LT |
822 | return 0; |
823 | ||
824 | need_lookup: | |
4ac91378 | 825 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
826 | if (IS_ERR(dentry)) |
827 | goto fail; | |
828 | goto done; | |
829 | ||
830 | need_revalidate: | |
bcdc5e01 IK |
831 | dentry = do_revalidate(dentry, nd); |
832 | if (!dentry) | |
833 | goto need_lookup; | |
834 | if (IS_ERR(dentry)) | |
835 | goto fail; | |
836 | goto done; | |
1da177e4 LT |
837 | |
838 | fail: | |
839 | return PTR_ERR(dentry); | |
840 | } | |
841 | ||
842 | /* | |
843 | * Name resolution. | |
ea3834d9 PM |
844 | * This is the basic name resolution function, turning a pathname into |
845 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 846 | * |
ea3834d9 PM |
847 | * Returns 0 and nd will have valid dentry and mnt on success. |
848 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 849 | */ |
fc9b52cd | 850 | static int __link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
851 | { |
852 | struct path next; | |
853 | struct inode *inode; | |
854 | int err; | |
855 | unsigned int lookup_flags = nd->flags; | |
856 | ||
857 | while (*name=='/') | |
858 | name++; | |
859 | if (!*name) | |
860 | goto return_reval; | |
861 | ||
4ac91378 | 862 | inode = nd->path.dentry->d_inode; |
1da177e4 | 863 | if (nd->depth) |
f55eab82 | 864 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
865 | |
866 | /* At this point we know we have a real path component. */ | |
867 | for(;;) { | |
868 | unsigned long hash; | |
869 | struct qstr this; | |
870 | unsigned int c; | |
871 | ||
cdce5d6b | 872 | nd->flags |= LOOKUP_CONTINUE; |
672b16b2 | 873 | err = exec_permission_lite(inode); |
e4543edd CH |
874 | if (err == -EAGAIN) |
875 | err = vfs_permission(nd, MAY_EXEC); | |
1da177e4 LT |
876 | if (err) |
877 | break; | |
878 | ||
879 | this.name = name; | |
880 | c = *(const unsigned char *)name; | |
881 | ||
882 | hash = init_name_hash(); | |
883 | do { | |
884 | name++; | |
885 | hash = partial_name_hash(c, hash); | |
886 | c = *(const unsigned char *)name; | |
887 | } while (c && (c != '/')); | |
888 | this.len = name - (const char *) this.name; | |
889 | this.hash = end_name_hash(hash); | |
890 | ||
891 | /* remove trailing slashes? */ | |
892 | if (!c) | |
893 | goto last_component; | |
894 | while (*++name == '/'); | |
895 | if (!*name) | |
896 | goto last_with_slashes; | |
897 | ||
898 | /* | |
899 | * "." and ".." are special - ".." especially so because it has | |
900 | * to be able to know about the current root directory and | |
901 | * parent relationships. | |
902 | */ | |
903 | if (this.name[0] == '.') switch (this.len) { | |
904 | default: | |
905 | break; | |
906 | case 2: | |
907 | if (this.name[1] != '.') | |
908 | break; | |
58c465eb | 909 | follow_dotdot(nd); |
4ac91378 | 910 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
911 | /* fallthrough */ |
912 | case 1: | |
913 | continue; | |
914 | } | |
915 | /* | |
916 | * See if the low-level filesystem might want | |
917 | * to use its own hash.. | |
918 | */ | |
4ac91378 JB |
919 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
920 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
921 | &this); | |
1da177e4 LT |
922 | if (err < 0) |
923 | break; | |
924 | } | |
1da177e4 LT |
925 | /* This does the actual lookups.. */ |
926 | err = do_lookup(nd, &this, &next); | |
927 | if (err) | |
928 | break; | |
1da177e4 LT |
929 | |
930 | err = -ENOENT; | |
931 | inode = next.dentry->d_inode; | |
932 | if (!inode) | |
933 | goto out_dput; | |
934 | err = -ENOTDIR; | |
935 | if (!inode->i_op) | |
936 | goto out_dput; | |
937 | ||
938 | if (inode->i_op->follow_link) { | |
90ebe565 | 939 | err = do_follow_link(&next, nd); |
1da177e4 LT |
940 | if (err) |
941 | goto return_err; | |
942 | err = -ENOENT; | |
4ac91378 | 943 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
944 | if (!inode) |
945 | break; | |
946 | err = -ENOTDIR; | |
947 | if (!inode->i_op) | |
948 | break; | |
09dd17d3 MS |
949 | } else |
950 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
951 | err = -ENOTDIR; |
952 | if (!inode->i_op->lookup) | |
953 | break; | |
954 | continue; | |
955 | /* here ends the main loop */ | |
956 | ||
957 | last_with_slashes: | |
958 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
959 | last_component: | |
f55eab82 TM |
960 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
961 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
962 | if (lookup_flags & LOOKUP_PARENT) |
963 | goto lookup_parent; | |
964 | if (this.name[0] == '.') switch (this.len) { | |
965 | default: | |
966 | break; | |
967 | case 2: | |
968 | if (this.name[1] != '.') | |
969 | break; | |
58c465eb | 970 | follow_dotdot(nd); |
4ac91378 | 971 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
972 | /* fallthrough */ |
973 | case 1: | |
974 | goto return_reval; | |
975 | } | |
4ac91378 JB |
976 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
977 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
978 | &this); | |
1da177e4 LT |
979 | if (err < 0) |
980 | break; | |
981 | } | |
982 | err = do_lookup(nd, &this, &next); | |
983 | if (err) | |
984 | break; | |
1da177e4 LT |
985 | inode = next.dentry->d_inode; |
986 | if ((lookup_flags & LOOKUP_FOLLOW) | |
987 | && inode && inode->i_op && inode->i_op->follow_link) { | |
90ebe565 | 988 | err = do_follow_link(&next, nd); |
1da177e4 LT |
989 | if (err) |
990 | goto return_err; | |
4ac91378 | 991 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
992 | } else |
993 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
994 | err = -ENOENT; |
995 | if (!inode) | |
996 | break; | |
997 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
998 | err = -ENOTDIR; | |
999 | if (!inode->i_op || !inode->i_op->lookup) | |
1000 | break; | |
1001 | } | |
1002 | goto return_base; | |
1003 | lookup_parent: | |
1004 | nd->last = this; | |
1005 | nd->last_type = LAST_NORM; | |
1006 | if (this.name[0] != '.') | |
1007 | goto return_base; | |
1008 | if (this.len == 1) | |
1009 | nd->last_type = LAST_DOT; | |
1010 | else if (this.len == 2 && this.name[1] == '.') | |
1011 | nd->last_type = LAST_DOTDOT; | |
1012 | else | |
1013 | goto return_base; | |
1014 | return_reval: | |
1015 | /* | |
1016 | * We bypassed the ordinary revalidation routines. | |
1017 | * We may need to check the cached dentry for staleness. | |
1018 | */ | |
4ac91378 JB |
1019 | if (nd->path.dentry && nd->path.dentry->d_sb && |
1020 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
1021 | err = -ESTALE; |
1022 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
1023 | if (!nd->path.dentry->d_op->d_revalidate( |
1024 | nd->path.dentry, nd)) | |
1da177e4 LT |
1025 | break; |
1026 | } | |
1027 | return_base: | |
1028 | return 0; | |
1029 | out_dput: | |
1d957f9b | 1030 | path_put_conditional(&next, nd); |
1da177e4 LT |
1031 | break; |
1032 | } | |
1d957f9b | 1033 | path_put(&nd->path); |
1da177e4 LT |
1034 | return_err: |
1035 | return err; | |
1036 | } | |
1037 | ||
fc9b52cd | 1038 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1039 | { |
1040 | current->total_link_count = 0; | |
1041 | return link_path_walk(name, nd); | |
1042 | } | |
1043 | ||
ea3834d9 | 1044 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
fc9b52cd | 1045 | static int do_path_lookup(int dfd, const char *name, |
5590ff0d | 1046 | unsigned int flags, struct nameidata *nd) |
1da177e4 | 1047 | { |
ea3834d9 | 1048 | int retval = 0; |
170aa3d0 UD |
1049 | int fput_needed; |
1050 | struct file *file; | |
e518ddb7 | 1051 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1052 | |
1053 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1054 | nd->flags = flags; | |
1055 | nd->depth = 0; | |
1056 | ||
1da177e4 | 1057 | if (*name=='/') { |
e518ddb7 | 1058 | read_lock(&fs->lock); |
6ac08c39 JB |
1059 | nd->path = fs->root; |
1060 | path_get(&fs->root); | |
e518ddb7 | 1061 | read_unlock(&fs->lock); |
5590ff0d | 1062 | } else if (dfd == AT_FDCWD) { |
e518ddb7 | 1063 | read_lock(&fs->lock); |
6ac08c39 JB |
1064 | nd->path = fs->pwd; |
1065 | path_get(&fs->pwd); | |
e518ddb7 | 1066 | read_unlock(&fs->lock); |
5590ff0d | 1067 | } else { |
5590ff0d UD |
1068 | struct dentry *dentry; |
1069 | ||
1070 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1071 | retval = -EBADF; |
1072 | if (!file) | |
6d09bb62 | 1073 | goto out_fail; |
5590ff0d | 1074 | |
0f7fc9e4 | 1075 | dentry = file->f_path.dentry; |
5590ff0d | 1076 | |
170aa3d0 UD |
1077 | retval = -ENOTDIR; |
1078 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1079 | goto fput_fail; |
5590ff0d UD |
1080 | |
1081 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1082 | if (retval) |
6d09bb62 | 1083 | goto fput_fail; |
5590ff0d | 1084 | |
5dd784d0 JB |
1085 | nd->path = file->f_path; |
1086 | path_get(&file->f_path); | |
5590ff0d UD |
1087 | |
1088 | fput_light(file, fput_needed); | |
1da177e4 | 1089 | } |
2dfdd266 JJS |
1090 | |
1091 | retval = path_walk(name, nd); | |
4ac91378 JB |
1092 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1093 | nd->path.dentry->d_inode)) | |
1094 | audit_inode(name, nd->path.dentry); | |
6d09bb62 | 1095 | out_fail: |
170aa3d0 UD |
1096 | return retval; |
1097 | ||
6d09bb62 | 1098 | fput_fail: |
170aa3d0 | 1099 | fput_light(file, fput_needed); |
6d09bb62 | 1100 | goto out_fail; |
1da177e4 LT |
1101 | } |
1102 | ||
fc9b52cd | 1103 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1104 | struct nameidata *nd) |
1105 | { | |
1106 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1107 | } | |
1108 | ||
d1811465 AV |
1109 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1110 | { | |
1111 | struct nameidata nd; | |
1112 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1113 | if (!res) | |
1114 | *path = nd.path; | |
1115 | return res; | |
1116 | } | |
1117 | ||
16f18200 JJS |
1118 | /** |
1119 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1120 | * @dentry: pointer to dentry of the base directory | |
1121 | * @mnt: pointer to vfs mount of the base directory | |
1122 | * @name: pointer to file name | |
1123 | * @flags: lookup flags | |
1124 | * @nd: pointer to nameidata | |
1125 | */ | |
1126 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1127 | const char *name, unsigned int flags, | |
1128 | struct nameidata *nd) | |
1129 | { | |
1130 | int retval; | |
1131 | ||
1132 | /* same as do_path_lookup */ | |
1133 | nd->last_type = LAST_ROOT; | |
1134 | nd->flags = flags; | |
1135 | nd->depth = 0; | |
1136 | ||
c8e7f449 JB |
1137 | nd->path.dentry = dentry; |
1138 | nd->path.mnt = mnt; | |
1139 | path_get(&nd->path); | |
16f18200 JJS |
1140 | |
1141 | retval = path_walk(name, nd); | |
4ac91378 JB |
1142 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1143 | nd->path.dentry->d_inode)) | |
1144 | audit_inode(name, nd->path.dentry); | |
16f18200 JJS |
1145 | |
1146 | return retval; | |
1147 | ||
1148 | } | |
1149 | ||
8737f3a1 AV |
1150 | /** |
1151 | * path_lookup_open - lookup a file path with open intent | |
1152 | * @dfd: the directory to use as base, or AT_FDCWD | |
1153 | * @name: pointer to file name | |
1154 | * @lookup_flags: lookup intent flags | |
1155 | * @nd: pointer to nameidata | |
1156 | * @open_flags: open intent flags | |
1157 | */ | |
1158 | int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags, | |
1159 | struct nameidata *nd, int open_flags) | |
834f2a4a TM |
1160 | { |
1161 | struct file *filp = get_empty_filp(); | |
1162 | int err; | |
1163 | ||
1164 | if (filp == NULL) | |
1165 | return -ENFILE; | |
1166 | nd->intent.open.file = filp; | |
1167 | nd->intent.open.flags = open_flags; | |
8737f3a1 | 1168 | nd->intent.open.create_mode = 0; |
5590ff0d | 1169 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1170 | if (IS_ERR(nd->intent.open.file)) { |
1171 | if (err == 0) { | |
1172 | err = PTR_ERR(nd->intent.open.file); | |
1d957f9b | 1173 | path_put(&nd->path); |
834f2a4a TM |
1174 | } |
1175 | } else if (err != 0) | |
1176 | release_open_intent(nd); | |
1177 | return err; | |
1178 | } | |
1179 | ||
eead1911 CH |
1180 | static struct dentry *__lookup_hash(struct qstr *name, |
1181 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1182 | { |
057f6c01 | 1183 | struct dentry *dentry; |
1da177e4 LT |
1184 | struct inode *inode; |
1185 | int err; | |
1186 | ||
1187 | inode = base->d_inode; | |
1da177e4 LT |
1188 | |
1189 | /* | |
1190 | * See if the low-level filesystem might want | |
1191 | * to use its own hash.. | |
1192 | */ | |
1193 | if (base->d_op && base->d_op->d_hash) { | |
1194 | err = base->d_op->d_hash(base, name); | |
1195 | dentry = ERR_PTR(err); | |
1196 | if (err < 0) | |
1197 | goto out; | |
1198 | } | |
1199 | ||
1200 | dentry = cached_lookup(base, name, nd); | |
1201 | if (!dentry) { | |
d70b67c8 MS |
1202 | struct dentry *new; |
1203 | ||
1204 | /* Don't create child dentry for a dead directory. */ | |
1205 | dentry = ERR_PTR(-ENOENT); | |
1206 | if (IS_DEADDIR(inode)) | |
1207 | goto out; | |
1208 | ||
1209 | new = d_alloc(base, name); | |
1da177e4 LT |
1210 | dentry = ERR_PTR(-ENOMEM); |
1211 | if (!new) | |
1212 | goto out; | |
1213 | dentry = inode->i_op->lookup(inode, new, nd); | |
1214 | if (!dentry) | |
1215 | dentry = new; | |
1216 | else | |
1217 | dput(new); | |
1218 | } | |
1219 | out: | |
1220 | return dentry; | |
1221 | } | |
1222 | ||
057f6c01 JM |
1223 | /* |
1224 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1225 | * needs parent already locked. Doesn't follow mounts. | |
1226 | * SMP-safe. | |
1227 | */ | |
eead1911 | 1228 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1229 | { |
057f6c01 JM |
1230 | int err; |
1231 | ||
f419a2e3 | 1232 | err = inode_permission(nd->path.dentry->d_inode, MAY_EXEC); |
057f6c01 | 1233 | if (err) |
eead1911 | 1234 | return ERR_PTR(err); |
4ac91378 | 1235 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1236 | } |
1237 | ||
eead1911 CH |
1238 | static int __lookup_one_len(const char *name, struct qstr *this, |
1239 | struct dentry *base, int len) | |
1da177e4 LT |
1240 | { |
1241 | unsigned long hash; | |
1da177e4 LT |
1242 | unsigned int c; |
1243 | ||
057f6c01 JM |
1244 | this->name = name; |
1245 | this->len = len; | |
1da177e4 | 1246 | if (!len) |
057f6c01 | 1247 | return -EACCES; |
1da177e4 LT |
1248 | |
1249 | hash = init_name_hash(); | |
1250 | while (len--) { | |
1251 | c = *(const unsigned char *)name++; | |
1252 | if (c == '/' || c == '\0') | |
057f6c01 | 1253 | return -EACCES; |
1da177e4 LT |
1254 | hash = partial_name_hash(c, hash); |
1255 | } | |
057f6c01 JM |
1256 | this->hash = end_name_hash(hash); |
1257 | return 0; | |
1258 | } | |
1da177e4 | 1259 | |
eead1911 | 1260 | /** |
a6b91919 | 1261 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1262 | * @name: pathname component to lookup |
1263 | * @base: base directory to lookup from | |
1264 | * @len: maximum length @len should be interpreted to | |
1265 | * | |
a6b91919 RD |
1266 | * Note that this routine is purely a helper for filesystem usage and should |
1267 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1268 | * nameidata argument is passed to the filesystem methods and a filesystem |
1269 | * using this helper needs to be prepared for that. | |
1270 | */ | |
057f6c01 JM |
1271 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1272 | { | |
1273 | int err; | |
1274 | struct qstr this; | |
1275 | ||
1276 | err = __lookup_one_len(name, &this, base, len); | |
eead1911 CH |
1277 | if (err) |
1278 | return ERR_PTR(err); | |
1279 | ||
f419a2e3 | 1280 | err = inode_permission(base->d_inode, MAY_EXEC); |
057f6c01 JM |
1281 | if (err) |
1282 | return ERR_PTR(err); | |
49705b77 | 1283 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1284 | } |
1285 | ||
eead1911 CH |
1286 | /** |
1287 | * lookup_one_noperm - bad hack for sysfs | |
1288 | * @name: pathname component to lookup | |
1289 | * @base: base directory to lookup from | |
1290 | * | |
1291 | * This is a variant of lookup_one_len that doesn't perform any permission | |
1292 | * checks. It's a horrible hack to work around the braindead sysfs | |
1293 | * architecture and should not be used anywhere else. | |
1294 | * | |
1295 | * DON'T USE THIS FUNCTION EVER, thanks. | |
1296 | */ | |
1297 | struct dentry *lookup_one_noperm(const char *name, struct dentry *base) | |
057f6c01 JM |
1298 | { |
1299 | int err; | |
1300 | struct qstr this; | |
1301 | ||
eead1911 | 1302 | err = __lookup_one_len(name, &this, base, strlen(name)); |
057f6c01 JM |
1303 | if (err) |
1304 | return ERR_PTR(err); | |
eead1911 | 1305 | return __lookup_hash(&this, base, NULL); |
1da177e4 LT |
1306 | } |
1307 | ||
2d8f3038 AV |
1308 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1309 | struct path *path) | |
1da177e4 | 1310 | { |
2d8f3038 | 1311 | struct nameidata nd; |
1da177e4 LT |
1312 | char *tmp = getname(name); |
1313 | int err = PTR_ERR(tmp); | |
1da177e4 | 1314 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1315 | |
1316 | BUG_ON(flags & LOOKUP_PARENT); | |
1317 | ||
1318 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1319 | putname(tmp); |
2d8f3038 AV |
1320 | if (!err) |
1321 | *path = nd.path; | |
1da177e4 LT |
1322 | } |
1323 | return err; | |
1324 | } | |
1325 | ||
2ad94ae6 AV |
1326 | static int user_path_parent(int dfd, const char __user *path, |
1327 | struct nameidata *nd, char **name) | |
1328 | { | |
1329 | char *s = getname(path); | |
1330 | int error; | |
1331 | ||
1332 | if (IS_ERR(s)) | |
1333 | return PTR_ERR(s); | |
1334 | ||
1335 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1336 | if (error) | |
1337 | putname(s); | |
1338 | else | |
1339 | *name = s; | |
1340 | ||
1341 | return error; | |
1342 | } | |
1343 | ||
1da177e4 LT |
1344 | /* |
1345 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1346 | * minimal. | |
1347 | */ | |
1348 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1349 | { | |
1350 | if (!(dir->i_mode & S_ISVTX)) | |
1351 | return 0; | |
1352 | if (inode->i_uid == current->fsuid) | |
1353 | return 0; | |
1354 | if (dir->i_uid == current->fsuid) | |
1355 | return 0; | |
1356 | return !capable(CAP_FOWNER); | |
1357 | } | |
1358 | ||
1359 | /* | |
1360 | * Check whether we can remove a link victim from directory dir, check | |
1361 | * whether the type of victim is right. | |
1362 | * 1. We can't do it if dir is read-only (done in permission()) | |
1363 | * 2. We should have write and exec permissions on dir | |
1364 | * 3. We can't remove anything from append-only dir | |
1365 | * 4. We can't do anything with immutable dir (done in permission()) | |
1366 | * 5. If the sticky bit on dir is set we should either | |
1367 | * a. be owner of dir, or | |
1368 | * b. be owner of victim, or | |
1369 | * c. have CAP_FOWNER capability | |
1370 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1371 | * links pointing to it. | |
1372 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1373 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1374 | * 9. We can't remove a root or mountpoint. | |
1375 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1376 | * nfs_async_unlink(). | |
1377 | */ | |
858119e1 | 1378 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1379 | { |
1380 | int error; | |
1381 | ||
1382 | if (!victim->d_inode) | |
1383 | return -ENOENT; | |
1384 | ||
1385 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1386 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1387 | |
f419a2e3 | 1388 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1389 | if (error) |
1390 | return error; | |
1391 | if (IS_APPEND(dir)) | |
1392 | return -EPERM; | |
1393 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
1394 | IS_IMMUTABLE(victim->d_inode)) | |
1395 | return -EPERM; | |
1396 | if (isdir) { | |
1397 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1398 | return -ENOTDIR; | |
1399 | if (IS_ROOT(victim)) | |
1400 | return -EBUSY; | |
1401 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1402 | return -EISDIR; | |
1403 | if (IS_DEADDIR(dir)) | |
1404 | return -ENOENT; | |
1405 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1406 | return -EBUSY; | |
1407 | return 0; | |
1408 | } | |
1409 | ||
1410 | /* Check whether we can create an object with dentry child in directory | |
1411 | * dir. | |
1412 | * 1. We can't do it if child already exists (open has special treatment for | |
1413 | * this case, but since we are inlined it's OK) | |
1414 | * 2. We can't do it if dir is read-only (done in permission()) | |
1415 | * 3. We should have write and exec permissions on dir | |
1416 | * 4. We can't do it if dir is immutable (done in permission()) | |
1417 | */ | |
a95164d9 | 1418 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1419 | { |
1420 | if (child->d_inode) | |
1421 | return -EEXIST; | |
1422 | if (IS_DEADDIR(dir)) | |
1423 | return -ENOENT; | |
f419a2e3 | 1424 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1425 | } |
1426 | ||
1427 | /* | |
1da177e4 LT |
1428 | * O_DIRECTORY translates into forcing a directory lookup. |
1429 | */ | |
1430 | static inline int lookup_flags(unsigned int f) | |
1431 | { | |
1432 | unsigned long retval = LOOKUP_FOLLOW; | |
1433 | ||
1434 | if (f & O_NOFOLLOW) | |
1435 | retval &= ~LOOKUP_FOLLOW; | |
1436 | ||
1da177e4 LT |
1437 | if (f & O_DIRECTORY) |
1438 | retval |= LOOKUP_DIRECTORY; | |
1439 | ||
1440 | return retval; | |
1441 | } | |
1442 | ||
1443 | /* | |
1444 | * p1 and p2 should be directories on the same fs. | |
1445 | */ | |
1446 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1447 | { | |
1448 | struct dentry *p; | |
1449 | ||
1450 | if (p1 == p2) { | |
f2eace23 | 1451 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1452 | return NULL; |
1453 | } | |
1454 | ||
a11f3a05 | 1455 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1456 | |
e2761a11 OH |
1457 | p = d_ancestor(p2, p1); |
1458 | if (p) { | |
1459 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1460 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1461 | return p; | |
1da177e4 LT |
1462 | } |
1463 | ||
e2761a11 OH |
1464 | p = d_ancestor(p1, p2); |
1465 | if (p) { | |
1466 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1467 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1468 | return p; | |
1da177e4 LT |
1469 | } |
1470 | ||
f2eace23 IM |
1471 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1472 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1473 | return NULL; |
1474 | } | |
1475 | ||
1476 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1477 | { | |
1b1dcc1b | 1478 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1479 | if (p1 != p2) { |
1b1dcc1b | 1480 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1481 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1482 | } |
1483 | } | |
1484 | ||
1485 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1486 | struct nameidata *nd) | |
1487 | { | |
a95164d9 | 1488 | int error = may_create(dir, dentry); |
1da177e4 LT |
1489 | |
1490 | if (error) | |
1491 | return error; | |
1492 | ||
1493 | if (!dir->i_op || !dir->i_op->create) | |
1494 | return -EACCES; /* shouldn't it be ENOSYS? */ | |
1495 | mode &= S_IALLUGO; | |
1496 | mode |= S_IFREG; | |
1497 | error = security_inode_create(dir, dentry, mode); | |
1498 | if (error) | |
1499 | return error; | |
1500 | DQUOT_INIT(dir); | |
1501 | error = dir->i_op->create(dir, dentry, mode, nd); | |
a74574aa | 1502 | if (!error) |
f38aa942 | 1503 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1504 | return error; |
1505 | } | |
1506 | ||
1507 | int may_open(struct nameidata *nd, int acc_mode, int flag) | |
1508 | { | |
4ac91378 | 1509 | struct dentry *dentry = nd->path.dentry; |
1da177e4 LT |
1510 | struct inode *inode = dentry->d_inode; |
1511 | int error; | |
1512 | ||
1513 | if (!inode) | |
1514 | return -ENOENT; | |
1515 | ||
1516 | if (S_ISLNK(inode->i_mode)) | |
1517 | return -ELOOP; | |
1518 | ||
974a9f0b | 1519 | if (S_ISDIR(inode->i_mode) && (acc_mode & MAY_WRITE)) |
1da177e4 LT |
1520 | return -EISDIR; |
1521 | ||
1da177e4 LT |
1522 | /* |
1523 | * FIFO's, sockets and device files are special: they don't | |
1524 | * actually live on the filesystem itself, and as such you | |
1525 | * can write to them even if the filesystem is read-only. | |
1526 | */ | |
1527 | if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { | |
1528 | flag &= ~O_TRUNC; | |
1529 | } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) { | |
4ac91378 | 1530 | if (nd->path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1531 | return -EACCES; |
1532 | ||
1533 | flag &= ~O_TRUNC; | |
4a3fd211 | 1534 | } |
b41572e9 DH |
1535 | |
1536 | error = vfs_permission(nd, acc_mode); | |
1537 | if (error) | |
1538 | return error; | |
1da177e4 LT |
1539 | /* |
1540 | * An append-only file must be opened in append mode for writing. | |
1541 | */ | |
1542 | if (IS_APPEND(inode)) { | |
1543 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
1544 | return -EPERM; | |
1545 | if (flag & O_TRUNC) | |
1546 | return -EPERM; | |
1547 | } | |
1548 | ||
1549 | /* O_NOATIME can only be set by the owner or superuser */ | |
1550 | if (flag & O_NOATIME) | |
3bd858ab | 1551 | if (!is_owner_or_cap(inode)) |
1da177e4 LT |
1552 | return -EPERM; |
1553 | ||
1554 | /* | |
1555 | * Ensure there are no outstanding leases on the file. | |
1556 | */ | |
1557 | error = break_lease(inode, flag); | |
1558 | if (error) | |
1559 | return error; | |
1560 | ||
1561 | if (flag & O_TRUNC) { | |
1562 | error = get_write_access(inode); | |
1563 | if (error) | |
1564 | return error; | |
1565 | ||
1566 | /* | |
1567 | * Refuse to truncate files with mandatory locks held on them. | |
1568 | */ | |
1569 | error = locks_verify_locked(inode); | |
1570 | if (!error) { | |
1571 | DQUOT_INIT(inode); | |
d139d7ff MS |
1572 | |
1573 | error = do_truncate(dentry, 0, | |
1574 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1575 | NULL); | |
1da177e4 LT |
1576 | } |
1577 | put_write_access(inode); | |
1578 | if (error) | |
1579 | return error; | |
1580 | } else | |
1581 | if (flag & FMODE_WRITE) | |
1582 | DQUOT_INIT(inode); | |
1583 | ||
1584 | return 0; | |
1585 | } | |
1586 | ||
d57999e1 DH |
1587 | /* |
1588 | * Be careful about ever adding any more callers of this | |
1589 | * function. Its flags must be in the namei format, not | |
1590 | * what get passed to sys_open(). | |
1591 | */ | |
1592 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1593 | int flag, int mode) |
1594 | { | |
1595 | int error; | |
4ac91378 | 1596 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1597 | |
1598 | if (!IS_POSIXACL(dir->d_inode)) | |
1599 | mode &= ~current->fs->umask; | |
1600 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); | |
1601 | mutex_unlock(&dir->d_inode->i_mutex); | |
4ac91378 JB |
1602 | dput(nd->path.dentry); |
1603 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1604 | if (error) |
1605 | return error; | |
1606 | /* Don't check for write permission, don't truncate */ | |
1607 | return may_open(nd, 0, flag & ~O_TRUNC); | |
1608 | } | |
1609 | ||
d57999e1 DH |
1610 | /* |
1611 | * Note that while the flag value (low two bits) for sys_open means: | |
1612 | * 00 - read-only | |
1613 | * 01 - write-only | |
1614 | * 10 - read-write | |
1615 | * 11 - special | |
1616 | * it is changed into | |
1617 | * 00 - no permissions needed | |
1618 | * 01 - read-permission | |
1619 | * 10 - write-permission | |
1620 | * 11 - read-write | |
1621 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1622 | * This is more logical, and also allows the 00 "no perm needed" | |
1623 | * to be used for symlinks (where the permissions are checked | |
1624 | * later). | |
1625 | * | |
1626 | */ | |
1627 | static inline int open_to_namei_flags(int flag) | |
1628 | { | |
1629 | if ((flag+1) & O_ACCMODE) | |
1630 | flag++; | |
1631 | return flag; | |
1632 | } | |
1633 | ||
4a3fd211 DH |
1634 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1635 | { | |
1636 | /* | |
1637 | * We'll never write to the fs underlying | |
1638 | * a device file. | |
1639 | */ | |
1640 | if (special_file(inode->i_mode)) | |
1641 | return 0; | |
1642 | return (flag & O_TRUNC); | |
1643 | } | |
1644 | ||
1da177e4 | 1645 | /* |
4a3fd211 DH |
1646 | * Note that the low bits of the passed in "open_flag" |
1647 | * are not the same as in the local variable "flag". See | |
1648 | * open_to_namei_flags() for more details. | |
1da177e4 | 1649 | */ |
a70e65df CH |
1650 | struct file *do_filp_open(int dfd, const char *pathname, |
1651 | int open_flag, int mode) | |
1da177e4 | 1652 | { |
4a3fd211 | 1653 | struct file *filp; |
a70e65df | 1654 | struct nameidata nd; |
834f2a4a | 1655 | int acc_mode, error; |
4e7506e4 | 1656 | struct path path; |
1da177e4 LT |
1657 | struct dentry *dir; |
1658 | int count = 0; | |
4a3fd211 | 1659 | int will_write; |
d57999e1 | 1660 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1661 | |
b77b0646 | 1662 | acc_mode = MAY_OPEN | ACC_MODE(flag); |
1da177e4 | 1663 | |
834f2a4a TM |
1664 | /* O_TRUNC implies we need access checks for write permissions */ |
1665 | if (flag & O_TRUNC) | |
1666 | acc_mode |= MAY_WRITE; | |
1667 | ||
1da177e4 LT |
1668 | /* Allow the LSM permission hook to distinguish append |
1669 | access from general write access. */ | |
1670 | if (flag & O_APPEND) | |
1671 | acc_mode |= MAY_APPEND; | |
1672 | ||
1da177e4 LT |
1673 | /* |
1674 | * The simplest case - just a plain lookup. | |
1675 | */ | |
1676 | if (!(flag & O_CREAT)) { | |
5590ff0d | 1677 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
a70e65df | 1678 | &nd, flag); |
1da177e4 | 1679 | if (error) |
a70e65df | 1680 | return ERR_PTR(error); |
1da177e4 LT |
1681 | goto ok; |
1682 | } | |
1683 | ||
1684 | /* | |
1685 | * Create - we need to know the parent. | |
1686 | */ | |
8737f3a1 | 1687 | error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1688 | if (error) |
a70e65df | 1689 | return ERR_PTR(error); |
1da177e4 LT |
1690 | |
1691 | /* | |
1692 | * We have the parent and last component. First of all, check | |
1693 | * that we are not asked to creat(2) an obvious directory - that | |
1694 | * will not do. | |
1695 | */ | |
1696 | error = -EISDIR; | |
a70e65df | 1697 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1698 | goto exit_parent; |
1da177e4 | 1699 | |
8737f3a1 AV |
1700 | error = -ENFILE; |
1701 | filp = get_empty_filp(); | |
1702 | if (filp == NULL) | |
1703 | goto exit_parent; | |
1704 | nd.intent.open.file = filp; | |
1705 | nd.intent.open.flags = flag; | |
1706 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1707 | dir = nd.path.dentry; |
1708 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1709 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1710 | if (flag & O_EXCL) |
1711 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1712 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1713 | path.dentry = lookup_hash(&nd); |
1714 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1715 | |
1716 | do_last: | |
4e7506e4 AV |
1717 | error = PTR_ERR(path.dentry); |
1718 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1719 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1720 | goto exit; |
1721 | } | |
1722 | ||
a70e65df | 1723 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1724 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1725 | goto exit_mutex_unlock; |
4af4c52f OD |
1726 | } |
1727 | ||
1da177e4 | 1728 | /* Negative dentry, just create the file */ |
4e7506e4 | 1729 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1730 | /* |
1731 | * This write is needed to ensure that a | |
1732 | * ro->rw transition does not occur between | |
1733 | * the time when the file is created and when | |
1734 | * a permanent write count is taken through | |
1735 | * the 'struct file' in nameidata_to_filp(). | |
1736 | */ | |
1737 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1738 | if (error) |
4a3fd211 DH |
1739 | goto exit_mutex_unlock; |
1740 | error = __open_namei_create(&nd, &path, flag, mode); | |
1741 | if (error) { | |
1742 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1743 | goto exit; |
4a3fd211 DH |
1744 | } |
1745 | filp = nameidata_to_filp(&nd, open_flag); | |
1746 | mnt_drop_write(nd.path.mnt); | |
1747 | return filp; | |
1da177e4 LT |
1748 | } |
1749 | ||
1750 | /* | |
1751 | * It already exists. | |
1752 | */ | |
1b1dcc1b | 1753 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1754 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1755 | |
1756 | error = -EEXIST; | |
1757 | if (flag & O_EXCL) | |
1758 | goto exit_dput; | |
1759 | ||
e13b210f | 1760 | if (__follow_mount(&path)) { |
1da177e4 | 1761 | error = -ELOOP; |
ba7a4c1a AV |
1762 | if (flag & O_NOFOLLOW) |
1763 | goto exit_dput; | |
1da177e4 | 1764 | } |
3e2efce0 | 1765 | |
1da177e4 | 1766 | error = -ENOENT; |
4e7506e4 | 1767 | if (!path.dentry->d_inode) |
1da177e4 | 1768 | goto exit_dput; |
4e7506e4 | 1769 | if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1770 | goto do_link; |
1771 | ||
a70e65df | 1772 | path_to_nameidata(&path, &nd); |
1da177e4 | 1773 | error = -EISDIR; |
4e7506e4 | 1774 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1775 | goto exit; |
1776 | ok: | |
4a3fd211 DH |
1777 | /* |
1778 | * Consider: | |
1779 | * 1. may_open() truncates a file | |
1780 | * 2. a rw->ro mount transition occurs | |
1781 | * 3. nameidata_to_filp() fails due to | |
1782 | * the ro mount. | |
1783 | * That would be inconsistent, and should | |
1784 | * be avoided. Taking this mnt write here | |
1785 | * ensures that (2) can not occur. | |
1786 | */ | |
1787 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1788 | if (will_write) { | |
1789 | error = mnt_want_write(nd.path.mnt); | |
1790 | if (error) | |
1791 | goto exit; | |
1792 | } | |
a70e65df | 1793 | error = may_open(&nd, acc_mode, flag); |
4a3fd211 DH |
1794 | if (error) { |
1795 | if (will_write) | |
1796 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1797 | goto exit; |
4a3fd211 DH |
1798 | } |
1799 | filp = nameidata_to_filp(&nd, open_flag); | |
1800 | /* | |
1801 | * It is now safe to drop the mnt write | |
1802 | * because the filp has had a write taken | |
1803 | * on its behalf. | |
1804 | */ | |
1805 | if (will_write) | |
1806 | mnt_drop_write(nd.path.mnt); | |
1807 | return filp; | |
1da177e4 | 1808 | |
4a3fd211 DH |
1809 | exit_mutex_unlock: |
1810 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1811 | exit_dput: |
a70e65df | 1812 | path_put_conditional(&path, &nd); |
1da177e4 | 1813 | exit: |
a70e65df CH |
1814 | if (!IS_ERR(nd.intent.open.file)) |
1815 | release_open_intent(&nd); | |
8737f3a1 | 1816 | exit_parent: |
a70e65df CH |
1817 | path_put(&nd.path); |
1818 | return ERR_PTR(error); | |
1da177e4 LT |
1819 | |
1820 | do_link: | |
1821 | error = -ELOOP; | |
1822 | if (flag & O_NOFOLLOW) | |
1823 | goto exit_dput; | |
1824 | /* | |
1825 | * This is subtle. Instead of calling do_follow_link() we do the | |
1826 | * thing by hands. The reason is that this way we have zero link_count | |
1827 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1828 | * After that we have the parent and last component, i.e. | |
1829 | * we are in the same situation as after the first path_walk(). | |
1830 | * Well, almost - if the last component is normal we get its copy | |
1831 | * stored in nd->last.name and we will have to putname() it when we | |
1832 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1833 | */ | |
a70e65df CH |
1834 | nd.flags |= LOOKUP_PARENT; |
1835 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1836 | if (error) |
1837 | goto exit_dput; | |
a70e65df | 1838 | error = __do_follow_link(&path, &nd); |
de459215 KK |
1839 | if (error) { |
1840 | /* Does someone understand code flow here? Or it is only | |
1841 | * me so stupid? Anathema to whoever designed this non-sense | |
1842 | * with "intent.open". | |
1843 | */ | |
a70e65df CH |
1844 | release_open_intent(&nd); |
1845 | return ERR_PTR(error); | |
de459215 | 1846 | } |
a70e65df CH |
1847 | nd.flags &= ~LOOKUP_PARENT; |
1848 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1849 | goto ok; |
1da177e4 | 1850 | error = -EISDIR; |
a70e65df | 1851 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1852 | goto exit; |
a70e65df CH |
1853 | if (nd.last.name[nd.last.len]) { |
1854 | __putname(nd.last.name); | |
1da177e4 LT |
1855 | goto exit; |
1856 | } | |
1857 | error = -ELOOP; | |
1858 | if (count++==32) { | |
a70e65df | 1859 | __putname(nd.last.name); |
1da177e4 LT |
1860 | goto exit; |
1861 | } | |
a70e65df | 1862 | dir = nd.path.dentry; |
1b1dcc1b | 1863 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1864 | path.dentry = lookup_hash(&nd); |
1865 | path.mnt = nd.path.mnt; | |
1866 | __putname(nd.last.name); | |
1da177e4 LT |
1867 | goto do_last; |
1868 | } | |
1869 | ||
a70e65df CH |
1870 | /** |
1871 | * filp_open - open file and return file pointer | |
1872 | * | |
1873 | * @filename: path to open | |
1874 | * @flags: open flags as per the open(2) second argument | |
1875 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1876 | * | |
1877 | * This is the helper to open a file from kernelspace if you really | |
1878 | * have to. But in generally you should not do this, so please move | |
1879 | * along, nothing to see here.. | |
1880 | */ | |
1881 | struct file *filp_open(const char *filename, int flags, int mode) | |
1882 | { | |
1883 | return do_filp_open(AT_FDCWD, filename, flags, mode); | |
1884 | } | |
1885 | EXPORT_SYMBOL(filp_open); | |
1886 | ||
1da177e4 LT |
1887 | /** |
1888 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1889 | * @nd: nameidata info | |
1890 | * @is_dir: directory flag | |
1891 | * | |
1892 | * Simple function to lookup and return a dentry and create it | |
1893 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1894 | * |
4ac91378 | 1895 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1896 | */ |
1897 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1898 | { | |
c663e5d8 | 1899 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1900 | |
4ac91378 | 1901 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1902 | /* |
1903 | * Yucky last component or no last component at all? | |
1904 | * (foo/., foo/.., /////) | |
1905 | */ | |
1da177e4 LT |
1906 | if (nd->last_type != LAST_NORM) |
1907 | goto fail; | |
1908 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1909 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1910 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1911 | |
1912 | /* | |
1913 | * Do the final lookup. | |
1914 | */ | |
49705b77 | 1915 | dentry = lookup_hash(nd); |
1da177e4 LT |
1916 | if (IS_ERR(dentry)) |
1917 | goto fail; | |
c663e5d8 | 1918 | |
e9baf6e5 AV |
1919 | if (dentry->d_inode) |
1920 | goto eexist; | |
c663e5d8 CH |
1921 | /* |
1922 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1923 | * From the vfs_mknod() POV we just have a negative dentry - | |
1924 | * all is fine. Let's be bastards - you had / on the end, you've | |
1925 | * been asking for (non-existent) directory. -ENOENT for you. | |
1926 | */ | |
e9baf6e5 AV |
1927 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1928 | dput(dentry); | |
1929 | dentry = ERR_PTR(-ENOENT); | |
1930 | } | |
1da177e4 | 1931 | return dentry; |
e9baf6e5 | 1932 | eexist: |
1da177e4 | 1933 | dput(dentry); |
e9baf6e5 | 1934 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1935 | fail: |
1936 | return dentry; | |
1937 | } | |
f81a0bff | 1938 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1939 | |
1940 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1941 | { | |
a95164d9 | 1942 | int error = may_create(dir, dentry); |
1da177e4 LT |
1943 | |
1944 | if (error) | |
1945 | return error; | |
1946 | ||
1947 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1948 | return -EPERM; | |
1949 | ||
1950 | if (!dir->i_op || !dir->i_op->mknod) | |
1951 | return -EPERM; | |
1952 | ||
08ce5f16 SH |
1953 | error = devcgroup_inode_mknod(mode, dev); |
1954 | if (error) | |
1955 | return error; | |
1956 | ||
1da177e4 LT |
1957 | error = security_inode_mknod(dir, dentry, mode, dev); |
1958 | if (error) | |
1959 | return error; | |
1960 | ||
1961 | DQUOT_INIT(dir); | |
1962 | error = dir->i_op->mknod(dir, dentry, mode, dev); | |
a74574aa | 1963 | if (!error) |
f38aa942 | 1964 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1965 | return error; |
1966 | } | |
1967 | ||
463c3197 DH |
1968 | static int may_mknod(mode_t mode) |
1969 | { | |
1970 | switch (mode & S_IFMT) { | |
1971 | case S_IFREG: | |
1972 | case S_IFCHR: | |
1973 | case S_IFBLK: | |
1974 | case S_IFIFO: | |
1975 | case S_IFSOCK: | |
1976 | case 0: /* zero mode translates to S_IFREG */ | |
1977 | return 0; | |
1978 | case S_IFDIR: | |
1979 | return -EPERM; | |
1980 | default: | |
1981 | return -EINVAL; | |
1982 | } | |
1983 | } | |
1984 | ||
5590ff0d UD |
1985 | asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode, |
1986 | unsigned dev) | |
1da177e4 | 1987 | { |
2ad94ae6 AV |
1988 | int error; |
1989 | char *tmp; | |
1990 | struct dentry *dentry; | |
1da177e4 LT |
1991 | struct nameidata nd; |
1992 | ||
1993 | if (S_ISDIR(mode)) | |
1994 | return -EPERM; | |
1da177e4 | 1995 | |
2ad94ae6 | 1996 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 1997 | if (error) |
2ad94ae6 AV |
1998 | return error; |
1999 | ||
1da177e4 | 2000 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2001 | if (IS_ERR(dentry)) { |
2002 | error = PTR_ERR(dentry); | |
2003 | goto out_unlock; | |
2004 | } | |
4ac91378 | 2005 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
1da177e4 | 2006 | mode &= ~current->fs->umask; |
463c3197 DH |
2007 | error = may_mknod(mode); |
2008 | if (error) | |
2009 | goto out_dput; | |
2010 | error = mnt_want_write(nd.path.mnt); | |
2011 | if (error) | |
2012 | goto out_dput; | |
2013 | switch (mode & S_IFMT) { | |
1da177e4 | 2014 | case 0: case S_IFREG: |
4ac91378 | 2015 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2016 | break; |
2017 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2018 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2019 | new_decode_dev(dev)); |
2020 | break; | |
2021 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2022 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2023 | break; |
1da177e4 | 2024 | } |
463c3197 DH |
2025 | mnt_drop_write(nd.path.mnt); |
2026 | out_dput: | |
2027 | dput(dentry); | |
2028 | out_unlock: | |
4ac91378 | 2029 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2030 | path_put(&nd.path); |
1da177e4 LT |
2031 | putname(tmp); |
2032 | ||
2033 | return error; | |
2034 | } | |
2035 | ||
5590ff0d UD |
2036 | asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev) |
2037 | { | |
2038 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2039 | } | |
2040 | ||
1da177e4 LT |
2041 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2042 | { | |
a95164d9 | 2043 | int error = may_create(dir, dentry); |
1da177e4 LT |
2044 | |
2045 | if (error) | |
2046 | return error; | |
2047 | ||
2048 | if (!dir->i_op || !dir->i_op->mkdir) | |
2049 | return -EPERM; | |
2050 | ||
2051 | mode &= (S_IRWXUGO|S_ISVTX); | |
2052 | error = security_inode_mkdir(dir, dentry, mode); | |
2053 | if (error) | |
2054 | return error; | |
2055 | ||
2056 | DQUOT_INIT(dir); | |
2057 | error = dir->i_op->mkdir(dir, dentry, mode); | |
a74574aa | 2058 | if (!error) |
f38aa942 | 2059 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2060 | return error; |
2061 | } | |
2062 | ||
5590ff0d | 2063 | asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode) |
1da177e4 LT |
2064 | { |
2065 | int error = 0; | |
2066 | char * tmp; | |
6902d925 DH |
2067 | struct dentry *dentry; |
2068 | struct nameidata nd; | |
1da177e4 | 2069 | |
2ad94ae6 AV |
2070 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2071 | if (error) | |
6902d925 | 2072 | goto out_err; |
1da177e4 | 2073 | |
6902d925 DH |
2074 | dentry = lookup_create(&nd, 1); |
2075 | error = PTR_ERR(dentry); | |
2076 | if (IS_ERR(dentry)) | |
2077 | goto out_unlock; | |
1da177e4 | 2078 | |
4ac91378 | 2079 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
6902d925 | 2080 | mode &= ~current->fs->umask; |
463c3197 DH |
2081 | error = mnt_want_write(nd.path.mnt); |
2082 | if (error) | |
2083 | goto out_dput; | |
4ac91378 | 2084 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
463c3197 DH |
2085 | mnt_drop_write(nd.path.mnt); |
2086 | out_dput: | |
6902d925 DH |
2087 | dput(dentry); |
2088 | out_unlock: | |
4ac91378 | 2089 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2090 | path_put(&nd.path); |
6902d925 DH |
2091 | putname(tmp); |
2092 | out_err: | |
1da177e4 LT |
2093 | return error; |
2094 | } | |
2095 | ||
5590ff0d UD |
2096 | asmlinkage long sys_mkdir(const char __user *pathname, int mode) |
2097 | { | |
2098 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2099 | } | |
2100 | ||
1da177e4 LT |
2101 | /* |
2102 | * We try to drop the dentry early: we should have | |
2103 | * a usage count of 2 if we're the only user of this | |
2104 | * dentry, and if that is true (possibly after pruning | |
2105 | * the dcache), then we drop the dentry now. | |
2106 | * | |
2107 | * A low-level filesystem can, if it choses, legally | |
2108 | * do a | |
2109 | * | |
2110 | * if (!d_unhashed(dentry)) | |
2111 | * return -EBUSY; | |
2112 | * | |
2113 | * if it cannot handle the case of removing a directory | |
2114 | * that is still in use by something else.. | |
2115 | */ | |
2116 | void dentry_unhash(struct dentry *dentry) | |
2117 | { | |
2118 | dget(dentry); | |
dc168427 | 2119 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2120 | spin_lock(&dcache_lock); |
2121 | spin_lock(&dentry->d_lock); | |
2122 | if (atomic_read(&dentry->d_count) == 2) | |
2123 | __d_drop(dentry); | |
2124 | spin_unlock(&dentry->d_lock); | |
2125 | spin_unlock(&dcache_lock); | |
2126 | } | |
2127 | ||
2128 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2129 | { | |
2130 | int error = may_delete(dir, dentry, 1); | |
2131 | ||
2132 | if (error) | |
2133 | return error; | |
2134 | ||
2135 | if (!dir->i_op || !dir->i_op->rmdir) | |
2136 | return -EPERM; | |
2137 | ||
2138 | DQUOT_INIT(dir); | |
2139 | ||
1b1dcc1b | 2140 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2141 | dentry_unhash(dentry); |
2142 | if (d_mountpoint(dentry)) | |
2143 | error = -EBUSY; | |
2144 | else { | |
2145 | error = security_inode_rmdir(dir, dentry); | |
2146 | if (!error) { | |
2147 | error = dir->i_op->rmdir(dir, dentry); | |
2148 | if (!error) | |
2149 | dentry->d_inode->i_flags |= S_DEAD; | |
2150 | } | |
2151 | } | |
1b1dcc1b | 2152 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2153 | if (!error) { |
1da177e4 LT |
2154 | d_delete(dentry); |
2155 | } | |
2156 | dput(dentry); | |
2157 | ||
2158 | return error; | |
2159 | } | |
2160 | ||
5590ff0d | 2161 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2162 | { |
2163 | int error = 0; | |
2164 | char * name; | |
2165 | struct dentry *dentry; | |
2166 | struct nameidata nd; | |
2167 | ||
2ad94ae6 | 2168 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2169 | if (error) |
2ad94ae6 | 2170 | return error; |
1da177e4 LT |
2171 | |
2172 | switch(nd.last_type) { | |
0612d9fb OH |
2173 | case LAST_DOTDOT: |
2174 | error = -ENOTEMPTY; | |
2175 | goto exit1; | |
2176 | case LAST_DOT: | |
2177 | error = -EINVAL; | |
2178 | goto exit1; | |
2179 | case LAST_ROOT: | |
2180 | error = -EBUSY; | |
2181 | goto exit1; | |
1da177e4 | 2182 | } |
0612d9fb OH |
2183 | |
2184 | nd.flags &= ~LOOKUP_PARENT; | |
2185 | ||
4ac91378 | 2186 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2187 | dentry = lookup_hash(&nd); |
1da177e4 | 2188 | error = PTR_ERR(dentry); |
6902d925 DH |
2189 | if (IS_ERR(dentry)) |
2190 | goto exit2; | |
0622753b DH |
2191 | error = mnt_want_write(nd.path.mnt); |
2192 | if (error) | |
2193 | goto exit3; | |
4ac91378 | 2194 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
0622753b DH |
2195 | mnt_drop_write(nd.path.mnt); |
2196 | exit3: | |
6902d925 DH |
2197 | dput(dentry); |
2198 | exit2: | |
4ac91378 | 2199 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2200 | exit1: |
1d957f9b | 2201 | path_put(&nd.path); |
1da177e4 LT |
2202 | putname(name); |
2203 | return error; | |
2204 | } | |
2205 | ||
5590ff0d UD |
2206 | asmlinkage long sys_rmdir(const char __user *pathname) |
2207 | { | |
2208 | return do_rmdir(AT_FDCWD, pathname); | |
2209 | } | |
2210 | ||
1da177e4 LT |
2211 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2212 | { | |
2213 | int error = may_delete(dir, dentry, 0); | |
2214 | ||
2215 | if (error) | |
2216 | return error; | |
2217 | ||
2218 | if (!dir->i_op || !dir->i_op->unlink) | |
2219 | return -EPERM; | |
2220 | ||
2221 | DQUOT_INIT(dir); | |
2222 | ||
1b1dcc1b | 2223 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2224 | if (d_mountpoint(dentry)) |
2225 | error = -EBUSY; | |
2226 | else { | |
2227 | error = security_inode_unlink(dir, dentry); | |
2228 | if (!error) | |
2229 | error = dir->i_op->unlink(dir, dentry); | |
2230 | } | |
1b1dcc1b | 2231 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2232 | |
2233 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2234 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2235 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2236 | d_delete(dentry); |
1da177e4 | 2237 | } |
0eeca283 | 2238 | |
1da177e4 LT |
2239 | return error; |
2240 | } | |
2241 | ||
2242 | /* | |
2243 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2244 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2245 | * writeout happening, and we don't want to prevent access to the directory |
2246 | * while waiting on the I/O. | |
2247 | */ | |
5590ff0d | 2248 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2249 | { |
2ad94ae6 AV |
2250 | int error; |
2251 | char *name; | |
1da177e4 LT |
2252 | struct dentry *dentry; |
2253 | struct nameidata nd; | |
2254 | struct inode *inode = NULL; | |
2255 | ||
2ad94ae6 | 2256 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2257 | if (error) |
2ad94ae6 AV |
2258 | return error; |
2259 | ||
1da177e4 LT |
2260 | error = -EISDIR; |
2261 | if (nd.last_type != LAST_NORM) | |
2262 | goto exit1; | |
0612d9fb OH |
2263 | |
2264 | nd.flags &= ~LOOKUP_PARENT; | |
2265 | ||
4ac91378 | 2266 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2267 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2268 | error = PTR_ERR(dentry); |
2269 | if (!IS_ERR(dentry)) { | |
2270 | /* Why not before? Because we want correct error value */ | |
2271 | if (nd.last.name[nd.last.len]) | |
2272 | goto slashes; | |
2273 | inode = dentry->d_inode; | |
2274 | if (inode) | |
2275 | atomic_inc(&inode->i_count); | |
0622753b DH |
2276 | error = mnt_want_write(nd.path.mnt); |
2277 | if (error) | |
2278 | goto exit2; | |
4ac91378 | 2279 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
0622753b | 2280 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2281 | exit2: |
2282 | dput(dentry); | |
2283 | } | |
4ac91378 | 2284 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2285 | if (inode) |
2286 | iput(inode); /* truncate the inode here */ | |
2287 | exit1: | |
1d957f9b | 2288 | path_put(&nd.path); |
1da177e4 LT |
2289 | putname(name); |
2290 | return error; | |
2291 | ||
2292 | slashes: | |
2293 | error = !dentry->d_inode ? -ENOENT : | |
2294 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2295 | goto exit2; | |
2296 | } | |
2297 | ||
5590ff0d UD |
2298 | asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag) |
2299 | { | |
2300 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2301 | return -EINVAL; | |
2302 | ||
2303 | if (flag & AT_REMOVEDIR) | |
2304 | return do_rmdir(dfd, pathname); | |
2305 | ||
2306 | return do_unlinkat(dfd, pathname); | |
2307 | } | |
2308 | ||
2309 | asmlinkage long sys_unlink(const char __user *pathname) | |
2310 | { | |
2311 | return do_unlinkat(AT_FDCWD, pathname); | |
2312 | } | |
2313 | ||
db2e747b | 2314 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2315 | { |
a95164d9 | 2316 | int error = may_create(dir, dentry); |
1da177e4 LT |
2317 | |
2318 | if (error) | |
2319 | return error; | |
2320 | ||
2321 | if (!dir->i_op || !dir->i_op->symlink) | |
2322 | return -EPERM; | |
2323 | ||
2324 | error = security_inode_symlink(dir, dentry, oldname); | |
2325 | if (error) | |
2326 | return error; | |
2327 | ||
2328 | DQUOT_INIT(dir); | |
2329 | error = dir->i_op->symlink(dir, dentry, oldname); | |
a74574aa | 2330 | if (!error) |
f38aa942 | 2331 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2332 | return error; |
2333 | } | |
2334 | ||
5590ff0d UD |
2335 | asmlinkage long sys_symlinkat(const char __user *oldname, |
2336 | int newdfd, const char __user *newname) | |
1da177e4 | 2337 | { |
2ad94ae6 AV |
2338 | int error; |
2339 | char *from; | |
2340 | char *to; | |
6902d925 DH |
2341 | struct dentry *dentry; |
2342 | struct nameidata nd; | |
1da177e4 LT |
2343 | |
2344 | from = getname(oldname); | |
2ad94ae6 | 2345 | if (IS_ERR(from)) |
1da177e4 | 2346 | return PTR_ERR(from); |
1da177e4 | 2347 | |
2ad94ae6 | 2348 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2349 | if (error) |
2ad94ae6 AV |
2350 | goto out_putname; |
2351 | ||
6902d925 DH |
2352 | dentry = lookup_create(&nd, 0); |
2353 | error = PTR_ERR(dentry); | |
2354 | if (IS_ERR(dentry)) | |
2355 | goto out_unlock; | |
2356 | ||
75c3f29d DH |
2357 | error = mnt_want_write(nd.path.mnt); |
2358 | if (error) | |
2359 | goto out_dput; | |
db2e747b | 2360 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
75c3f29d DH |
2361 | mnt_drop_write(nd.path.mnt); |
2362 | out_dput: | |
6902d925 DH |
2363 | dput(dentry); |
2364 | out_unlock: | |
4ac91378 | 2365 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2366 | path_put(&nd.path); |
6902d925 DH |
2367 | putname(to); |
2368 | out_putname: | |
1da177e4 LT |
2369 | putname(from); |
2370 | return error; | |
2371 | } | |
2372 | ||
5590ff0d UD |
2373 | asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname) |
2374 | { | |
2375 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2376 | } | |
2377 | ||
1da177e4 LT |
2378 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2379 | { | |
2380 | struct inode *inode = old_dentry->d_inode; | |
2381 | int error; | |
2382 | ||
2383 | if (!inode) | |
2384 | return -ENOENT; | |
2385 | ||
a95164d9 | 2386 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2387 | if (error) |
2388 | return error; | |
2389 | ||
2390 | if (dir->i_sb != inode->i_sb) | |
2391 | return -EXDEV; | |
2392 | ||
2393 | /* | |
2394 | * A link to an append-only or immutable file cannot be created. | |
2395 | */ | |
2396 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2397 | return -EPERM; | |
2398 | if (!dir->i_op || !dir->i_op->link) | |
2399 | return -EPERM; | |
7e79eedb | 2400 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2401 | return -EPERM; |
2402 | ||
2403 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2404 | if (error) | |
2405 | return error; | |
2406 | ||
7e79eedb | 2407 | mutex_lock(&inode->i_mutex); |
1da177e4 LT |
2408 | DQUOT_INIT(dir); |
2409 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 2410 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2411 | if (!error) |
7e79eedb | 2412 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2413 | return error; |
2414 | } | |
2415 | ||
2416 | /* | |
2417 | * Hardlinks are often used in delicate situations. We avoid | |
2418 | * security-related surprises by not following symlinks on the | |
2419 | * newname. --KAB | |
2420 | * | |
2421 | * We don't follow them on the oldname either to be compatible | |
2422 | * with linux 2.0, and to avoid hard-linking to directories | |
2423 | * and other special files. --ADM | |
2424 | */ | |
5590ff0d | 2425 | asmlinkage long sys_linkat(int olddfd, const char __user *oldname, |
c04030e1 UD |
2426 | int newdfd, const char __user *newname, |
2427 | int flags) | |
1da177e4 LT |
2428 | { |
2429 | struct dentry *new_dentry; | |
2d8f3038 AV |
2430 | struct nameidata nd; |
2431 | struct path old_path; | |
1da177e4 | 2432 | int error; |
2ad94ae6 | 2433 | char *to; |
1da177e4 | 2434 | |
45c9b11a | 2435 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2436 | return -EINVAL; |
2437 | ||
2d8f3038 AV |
2438 | error = user_path_at(olddfd, oldname, |
2439 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2440 | &old_path); | |
1da177e4 | 2441 | if (error) |
2ad94ae6 AV |
2442 | return error; |
2443 | ||
2444 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2445 | if (error) |
2446 | goto out; | |
2447 | error = -EXDEV; | |
2d8f3038 | 2448 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2449 | goto out_release; |
2450 | new_dentry = lookup_create(&nd, 0); | |
2451 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2452 | if (IS_ERR(new_dentry)) |
2453 | goto out_unlock; | |
75c3f29d DH |
2454 | error = mnt_want_write(nd.path.mnt); |
2455 | if (error) | |
2456 | goto out_dput; | |
2d8f3038 | 2457 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
75c3f29d DH |
2458 | mnt_drop_write(nd.path.mnt); |
2459 | out_dput: | |
6902d925 DH |
2460 | dput(new_dentry); |
2461 | out_unlock: | |
4ac91378 | 2462 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2463 | out_release: |
1d957f9b | 2464 | path_put(&nd.path); |
2ad94ae6 | 2465 | putname(to); |
1da177e4 | 2466 | out: |
2d8f3038 | 2467 | path_put(&old_path); |
1da177e4 LT |
2468 | |
2469 | return error; | |
2470 | } | |
2471 | ||
5590ff0d UD |
2472 | asmlinkage long sys_link(const char __user *oldname, const char __user *newname) |
2473 | { | |
c04030e1 | 2474 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2475 | } |
2476 | ||
1da177e4 LT |
2477 | /* |
2478 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2479 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2480 | * Problems: | |
2481 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2482 | * b) race potential - two innocent renames can create a loop together. | |
2483 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2484 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2485 | * story. |
2486 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2487 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2488 | * whether the target exists). Solution: try to be smart with locking |
2489 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2490 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2491 | * move will be locked. Thus we can rank directories by the tree |
2492 | * (ancestors first) and rank all non-directories after them. | |
2493 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2494 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2495 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2496 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2497 | * we'd better make sure that there's no link(2) for them. | |
2498 | * d) some filesystems don't support opened-but-unlinked directories, | |
2499 | * either because of layout or because they are not ready to deal with | |
2500 | * all cases correctly. The latter will be fixed (taking this sort of | |
2501 | * stuff into VFS), but the former is not going away. Solution: the same | |
2502 | * trick as in rmdir(). | |
2503 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2504 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2505 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2506 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2507 | * locking]. |
2508 | */ | |
75c96f85 AB |
2509 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2510 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2511 | { |
2512 | int error = 0; | |
2513 | struct inode *target; | |
2514 | ||
2515 | /* | |
2516 | * If we are going to change the parent - check write permissions, | |
2517 | * we'll need to flip '..'. | |
2518 | */ | |
2519 | if (new_dir != old_dir) { | |
f419a2e3 | 2520 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2521 | if (error) |
2522 | return error; | |
2523 | } | |
2524 | ||
2525 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2526 | if (error) | |
2527 | return error; | |
2528 | ||
2529 | target = new_dentry->d_inode; | |
2530 | if (target) { | |
1b1dcc1b | 2531 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2532 | dentry_unhash(new_dentry); |
2533 | } | |
2534 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2535 | error = -EBUSY; | |
2536 | else | |
2537 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2538 | if (target) { | |
2539 | if (!error) | |
2540 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2541 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2542 | if (d_unhashed(new_dentry)) |
2543 | d_rehash(new_dentry); | |
2544 | dput(new_dentry); | |
2545 | } | |
e31e14ec | 2546 | if (!error) |
349457cc MF |
2547 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2548 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2549 | return error; |
2550 | } | |
2551 | ||
75c96f85 AB |
2552 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2553 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2554 | { |
2555 | struct inode *target; | |
2556 | int error; | |
2557 | ||
2558 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2559 | if (error) | |
2560 | return error; | |
2561 | ||
2562 | dget(new_dentry); | |
2563 | target = new_dentry->d_inode; | |
2564 | if (target) | |
1b1dcc1b | 2565 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2566 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2567 | error = -EBUSY; | |
2568 | else | |
2569 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2570 | if (!error) { | |
349457cc | 2571 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2572 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2573 | } |
2574 | if (target) | |
1b1dcc1b | 2575 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2576 | dput(new_dentry); |
2577 | return error; | |
2578 | } | |
2579 | ||
2580 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2581 | struct inode *new_dir, struct dentry *new_dentry) | |
2582 | { | |
2583 | int error; | |
2584 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2585 | const char *old_name; |
1da177e4 LT |
2586 | |
2587 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2588 | return 0; | |
2589 | ||
2590 | error = may_delete(old_dir, old_dentry, is_dir); | |
2591 | if (error) | |
2592 | return error; | |
2593 | ||
2594 | if (!new_dentry->d_inode) | |
a95164d9 | 2595 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2596 | else |
2597 | error = may_delete(new_dir, new_dentry, is_dir); | |
2598 | if (error) | |
2599 | return error; | |
2600 | ||
2601 | if (!old_dir->i_op || !old_dir->i_op->rename) | |
2602 | return -EPERM; | |
2603 | ||
2604 | DQUOT_INIT(old_dir); | |
2605 | DQUOT_INIT(new_dir); | |
2606 | ||
0eeca283 RL |
2607 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2608 | ||
1da177e4 LT |
2609 | if (is_dir) |
2610 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2611 | else | |
2612 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2613 | if (!error) { | |
0eeca283 | 2614 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2615 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2616 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2617 | } |
0eeca283 RL |
2618 | fsnotify_oldname_free(old_name); |
2619 | ||
1da177e4 LT |
2620 | return error; |
2621 | } | |
2622 | ||
2ad94ae6 AV |
2623 | asmlinkage long sys_renameat(int olddfd, const char __user *oldname, |
2624 | int newdfd, const char __user *newname) | |
1da177e4 | 2625 | { |
2ad94ae6 AV |
2626 | struct dentry *old_dir, *new_dir; |
2627 | struct dentry *old_dentry, *new_dentry; | |
2628 | struct dentry *trap; | |
1da177e4 | 2629 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2630 | char *from; |
2631 | char *to; | |
2632 | int error; | |
1da177e4 | 2633 | |
2ad94ae6 | 2634 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2635 | if (error) |
2636 | goto exit; | |
2637 | ||
2ad94ae6 | 2638 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2639 | if (error) |
2640 | goto exit1; | |
2641 | ||
2642 | error = -EXDEV; | |
4ac91378 | 2643 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2644 | goto exit2; |
2645 | ||
4ac91378 | 2646 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2647 | error = -EBUSY; |
2648 | if (oldnd.last_type != LAST_NORM) | |
2649 | goto exit2; | |
2650 | ||
4ac91378 | 2651 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2652 | if (newnd.last_type != LAST_NORM) |
2653 | goto exit2; | |
2654 | ||
0612d9fb OH |
2655 | oldnd.flags &= ~LOOKUP_PARENT; |
2656 | newnd.flags &= ~LOOKUP_PARENT; | |
2657 | ||
1da177e4 LT |
2658 | trap = lock_rename(new_dir, old_dir); |
2659 | ||
49705b77 | 2660 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2661 | error = PTR_ERR(old_dentry); |
2662 | if (IS_ERR(old_dentry)) | |
2663 | goto exit3; | |
2664 | /* source must exist */ | |
2665 | error = -ENOENT; | |
2666 | if (!old_dentry->d_inode) | |
2667 | goto exit4; | |
2668 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2669 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2670 | error = -ENOTDIR; | |
2671 | if (oldnd.last.name[oldnd.last.len]) | |
2672 | goto exit4; | |
2673 | if (newnd.last.name[newnd.last.len]) | |
2674 | goto exit4; | |
2675 | } | |
2676 | /* source should not be ancestor of target */ | |
2677 | error = -EINVAL; | |
2678 | if (old_dentry == trap) | |
2679 | goto exit4; | |
49705b77 | 2680 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2681 | error = PTR_ERR(new_dentry); |
2682 | if (IS_ERR(new_dentry)) | |
2683 | goto exit4; | |
2684 | /* target should not be an ancestor of source */ | |
2685 | error = -ENOTEMPTY; | |
2686 | if (new_dentry == trap) | |
2687 | goto exit5; | |
2688 | ||
9079b1eb DH |
2689 | error = mnt_want_write(oldnd.path.mnt); |
2690 | if (error) | |
2691 | goto exit5; | |
1da177e4 LT |
2692 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2693 | new_dir->d_inode, new_dentry); | |
9079b1eb | 2694 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2695 | exit5: |
2696 | dput(new_dentry); | |
2697 | exit4: | |
2698 | dput(old_dentry); | |
2699 | exit3: | |
2700 | unlock_rename(new_dir, old_dir); | |
2701 | exit2: | |
1d957f9b | 2702 | path_put(&newnd.path); |
2ad94ae6 | 2703 | putname(to); |
1da177e4 | 2704 | exit1: |
1d957f9b | 2705 | path_put(&oldnd.path); |
1da177e4 | 2706 | putname(from); |
2ad94ae6 | 2707 | exit: |
1da177e4 LT |
2708 | return error; |
2709 | } | |
2710 | ||
5590ff0d UD |
2711 | asmlinkage long sys_rename(const char __user *oldname, const char __user *newname) |
2712 | { | |
2713 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2714 | } | |
2715 | ||
1da177e4 LT |
2716 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2717 | { | |
2718 | int len; | |
2719 | ||
2720 | len = PTR_ERR(link); | |
2721 | if (IS_ERR(link)) | |
2722 | goto out; | |
2723 | ||
2724 | len = strlen(link); | |
2725 | if (len > (unsigned) buflen) | |
2726 | len = buflen; | |
2727 | if (copy_to_user(buffer, link, len)) | |
2728 | len = -EFAULT; | |
2729 | out: | |
2730 | return len; | |
2731 | } | |
2732 | ||
2733 | /* | |
2734 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2735 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2736 | * using) it for any given inode is up to filesystem. | |
2737 | */ | |
2738 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2739 | { | |
2740 | struct nameidata nd; | |
cc314eef | 2741 | void *cookie; |
694a1764 | 2742 | int res; |
cc314eef | 2743 | |
1da177e4 | 2744 | nd.depth = 0; |
cc314eef | 2745 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2746 | if (IS_ERR(cookie)) |
2747 | return PTR_ERR(cookie); | |
2748 | ||
2749 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2750 | if (dentry->d_inode->i_op->put_link) | |
2751 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2752 | return res; | |
1da177e4 LT |
2753 | } |
2754 | ||
2755 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2756 | { | |
2757 | return __vfs_follow_link(nd, link); | |
2758 | } | |
2759 | ||
2760 | /* get the link contents into pagecache */ | |
2761 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2762 | { | |
2763 | struct page * page; | |
2764 | struct address_space *mapping = dentry->d_inode->i_mapping; | |
090d2b18 | 2765 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2766 | if (IS_ERR(page)) |
6fe6900e | 2767 | return (char*)page; |
1da177e4 LT |
2768 | *ppage = page; |
2769 | return kmap(page); | |
1da177e4 LT |
2770 | } |
2771 | ||
2772 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2773 | { | |
2774 | struct page *page = NULL; | |
2775 | char *s = page_getlink(dentry, &page); | |
2776 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2777 | if (page) { | |
2778 | kunmap(page); | |
2779 | page_cache_release(page); | |
2780 | } | |
2781 | return res; | |
2782 | } | |
2783 | ||
cc314eef | 2784 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2785 | { |
cc314eef | 2786 | struct page *page = NULL; |
1da177e4 | 2787 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2788 | return page; |
1da177e4 LT |
2789 | } |
2790 | ||
cc314eef | 2791 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2792 | { |
cc314eef LT |
2793 | struct page *page = cookie; |
2794 | ||
2795 | if (page) { | |
1da177e4 LT |
2796 | kunmap(page); |
2797 | page_cache_release(page); | |
1da177e4 LT |
2798 | } |
2799 | } | |
2800 | ||
0adb25d2 KK |
2801 | int __page_symlink(struct inode *inode, const char *symname, int len, |
2802 | gfp_t gfp_mask) | |
1da177e4 LT |
2803 | { |
2804 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2805 | struct page *page; |
afddba49 | 2806 | void *fsdata; |
beb497ab | 2807 | int err; |
1da177e4 LT |
2808 | char *kaddr; |
2809 | ||
7e53cac4 | 2810 | retry: |
afddba49 NP |
2811 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
2812 | AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata); | |
1da177e4 | 2813 | if (err) |
afddba49 NP |
2814 | goto fail; |
2815 | ||
1da177e4 LT |
2816 | kaddr = kmap_atomic(page, KM_USER0); |
2817 | memcpy(kaddr, symname, len-1); | |
2818 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2819 | |
2820 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2821 | page, fsdata); | |
1da177e4 LT |
2822 | if (err < 0) |
2823 | goto fail; | |
afddba49 NP |
2824 | if (err < len-1) |
2825 | goto retry; | |
2826 | ||
1da177e4 LT |
2827 | mark_inode_dirty(inode); |
2828 | return 0; | |
1da177e4 LT |
2829 | fail: |
2830 | return err; | |
2831 | } | |
2832 | ||
0adb25d2 KK |
2833 | int page_symlink(struct inode *inode, const char *symname, int len) |
2834 | { | |
2835 | return __page_symlink(inode, symname, len, | |
2836 | mapping_gfp_mask(inode->i_mapping)); | |
2837 | } | |
2838 | ||
92e1d5be | 2839 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2840 | .readlink = generic_readlink, |
2841 | .follow_link = page_follow_link_light, | |
2842 | .put_link = page_put_link, | |
2843 | }; | |
2844 | ||
2d8f3038 | 2845 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2846 | EXPORT_SYMBOL(follow_down); |
2847 | EXPORT_SYMBOL(follow_up); | |
2848 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2849 | EXPORT_SYMBOL(getname); | |
2850 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2851 | EXPORT_SYMBOL(lookup_one_len); |
2852 | EXPORT_SYMBOL(page_follow_link_light); | |
2853 | EXPORT_SYMBOL(page_put_link); | |
2854 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2855 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2856 | EXPORT_SYMBOL(page_symlink); |
2857 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2858 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2859 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2860 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2861 | EXPORT_SYMBOL(inode_permission); |
e4543edd | 2862 | EXPORT_SYMBOL(vfs_permission); |
8c744fb8 | 2863 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2864 | EXPORT_SYMBOL(unlock_rename); |
2865 | EXPORT_SYMBOL(vfs_create); | |
2866 | EXPORT_SYMBOL(vfs_follow_link); | |
2867 | EXPORT_SYMBOL(vfs_link); | |
2868 | EXPORT_SYMBOL(vfs_mkdir); | |
2869 | EXPORT_SYMBOL(vfs_mknod); | |
2870 | EXPORT_SYMBOL(generic_permission); | |
2871 | EXPORT_SYMBOL(vfs_readlink); | |
2872 | EXPORT_SYMBOL(vfs_rename); | |
2873 | EXPORT_SYMBOL(vfs_rmdir); | |
2874 | EXPORT_SYMBOL(vfs_symlink); | |
2875 | EXPORT_SYMBOL(vfs_unlink); | |
2876 | EXPORT_SYMBOL(dentry_unhash); | |
2877 | EXPORT_SYMBOL(generic_readlink); |