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/namei.h> |
35 | #include <asm/uaccess.h> | |
36 | ||
37 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
38 | ||
39 | /* [Feb-1997 T. Schoebel-Theuer] | |
40 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
41 | * were necessary because of omirr. The reason is that omirr needs | |
42 | * to know the _real_ pathname, not the user-supplied one, in case | |
43 | * of symlinks (and also when transname replacements occur). | |
44 | * | |
45 | * The new code replaces the old recursive symlink resolution with | |
46 | * an iterative one (in case of non-nested symlink chains). It does | |
47 | * this with calls to <fs>_follow_link(). | |
48 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
49 | * replaced with a single function lookup_dentry() that can handle all | |
50 | * the special cases of the former code. | |
51 | * | |
52 | * With the new dcache, the pathname is stored at each inode, at least as | |
53 | * long as the refcount of the inode is positive. As a side effect, the | |
54 | * size of the dcache depends on the inode cache and thus is dynamic. | |
55 | * | |
56 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
57 | * resolution to correspond with current state of the code. | |
58 | * | |
59 | * Note that the symlink resolution is not *completely* iterative. | |
60 | * There is still a significant amount of tail- and mid- recursion in | |
61 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
62 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
63 | * may return different results than <fs>_follow_link(). Many virtual | |
64 | * filesystems (including /proc) exhibit this behavior. | |
65 | */ | |
66 | ||
67 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
68 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
69 | * and the name already exists in form of a symlink, try to create the new | |
70 | * name indicated by the symlink. The old code always complained that the | |
71 | * name already exists, due to not following the symlink even if its target | |
72 | * is nonexistent. The new semantics affects also mknod() and link() when | |
73 | * the name is a symlink pointing to a non-existant name. | |
74 | * | |
75 | * I don't know which semantics is the right one, since I have no access | |
76 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
77 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
78 | * "old" one. Personally, I think the new semantics is much more logical. | |
79 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
80 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
81 | * and in the old Linux semantics. | |
82 | */ | |
83 | ||
84 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
85 | * semantics. See the comments in "open_namei" and "do_link" below. | |
86 | * | |
87 | * [10-Sep-98 Alan Modra] Another symlink change. | |
88 | */ | |
89 | ||
90 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
91 | * inside the path - always follow. | |
92 | * in the last component in creation/removal/renaming - never follow. | |
93 | * if LOOKUP_FOLLOW passed - follow. | |
94 | * if the pathname has trailing slashes - follow. | |
95 | * otherwise - don't follow. | |
96 | * (applied in that order). | |
97 | * | |
98 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
99 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
100 | * During the 2.4 we need to fix the userland stuff depending on it - | |
101 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
102 | * XEmacs seems to be relying on it... | |
103 | */ | |
104 | /* | |
105 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 106 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
107 | * any extra contention... |
108 | */ | |
109 | ||
a02f76c3 | 110 | static int __link_path_walk(const char *name, struct nameidata *nd); |
c4a7808f | 111 | |
1da177e4 LT |
112 | /* In order to reduce some races, while at the same time doing additional |
113 | * checking and hopefully speeding things up, we copy filenames to the | |
114 | * kernel data space before using them.. | |
115 | * | |
116 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
117 | * PATH_MAX includes the nul terminator --RR. | |
118 | */ | |
858119e1 | 119 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
120 | { |
121 | int retval; | |
122 | unsigned long len = PATH_MAX; | |
123 | ||
124 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
125 | if ((unsigned long) filename >= TASK_SIZE) | |
126 | return -EFAULT; | |
127 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
128 | len = TASK_SIZE - (unsigned long) filename; | |
129 | } | |
130 | ||
131 | retval = strncpy_from_user(page, filename, len); | |
132 | if (retval > 0) { | |
133 | if (retval < len) | |
134 | return 0; | |
135 | return -ENAMETOOLONG; | |
136 | } else if (!retval) | |
137 | retval = -ENOENT; | |
138 | return retval; | |
139 | } | |
140 | ||
141 | char * getname(const char __user * filename) | |
142 | { | |
143 | char *tmp, *result; | |
144 | ||
145 | result = ERR_PTR(-ENOMEM); | |
146 | tmp = __getname(); | |
147 | if (tmp) { | |
148 | int retval = do_getname(filename, tmp); | |
149 | ||
150 | result = tmp; | |
151 | if (retval < 0) { | |
152 | __putname(tmp); | |
153 | result = ERR_PTR(retval); | |
154 | } | |
155 | } | |
156 | audit_getname(result); | |
157 | return result; | |
158 | } | |
159 | ||
160 | #ifdef CONFIG_AUDITSYSCALL | |
161 | void putname(const char *name) | |
162 | { | |
5ac3a9c2 | 163 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
164 | audit_putname(name); |
165 | else | |
166 | __putname(name); | |
167 | } | |
168 | EXPORT_SYMBOL(putname); | |
169 | #endif | |
170 | ||
171 | ||
172 | /** | |
173 | * generic_permission - check for access rights on a Posix-like filesystem | |
174 | * @inode: inode to check access rights for | |
175 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
176 | * @check_acl: optional callback to check for Posix ACLs | |
177 | * | |
178 | * Used to check for read/write/execute permissions on a file. | |
179 | * We use "fsuid" for this, letting us set arbitrary permissions | |
180 | * for filesystem access without changing the "normal" uids which | |
181 | * are used for other things.. | |
182 | */ | |
183 | int generic_permission(struct inode *inode, int mask, | |
184 | int (*check_acl)(struct inode *inode, int mask)) | |
185 | { | |
186 | umode_t mode = inode->i_mode; | |
187 | ||
188 | if (current->fsuid == inode->i_uid) | |
189 | mode >>= 6; | |
190 | else { | |
191 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
192 | int error = check_acl(inode, mask); | |
193 | if (error == -EACCES) | |
194 | goto check_capabilities; | |
195 | else if (error != -EAGAIN) | |
196 | return error; | |
197 | } | |
198 | ||
199 | if (in_group_p(inode->i_gid)) | |
200 | mode >>= 3; | |
201 | } | |
202 | ||
203 | /* | |
204 | * If the DACs are ok we don't need any capability check. | |
205 | */ | |
206 | if (((mode & mask & (MAY_READ|MAY_WRITE|MAY_EXEC)) == mask)) | |
207 | return 0; | |
208 | ||
209 | check_capabilities: | |
210 | /* | |
211 | * Read/write DACs are always overridable. | |
212 | * Executable DACs are overridable if at least one exec bit is set. | |
213 | */ | |
214 | if (!(mask & MAY_EXEC) || | |
215 | (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode)) | |
216 | if (capable(CAP_DAC_OVERRIDE)) | |
217 | return 0; | |
218 | ||
219 | /* | |
220 | * Searching includes executable on directories, else just read. | |
221 | */ | |
222 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
223 | if (capable(CAP_DAC_READ_SEARCH)) | |
224 | return 0; | |
225 | ||
226 | return -EACCES; | |
227 | } | |
228 | ||
229 | int permission(struct inode *inode, int mask, struct nameidata *nd) | |
230 | { | |
231 | int retval, submask; | |
c7eb2667 DH |
232 | struct vfsmount *mnt = NULL; |
233 | ||
234 | if (nd) | |
4ac91378 | 235 | mnt = nd->path.mnt; |
1da177e4 LT |
236 | |
237 | if (mask & MAY_WRITE) { | |
22590e41 | 238 | umode_t mode = inode->i_mode; |
1da177e4 LT |
239 | |
240 | /* | |
241 | * Nobody gets write access to a read-only fs. | |
242 | */ | |
243 | if (IS_RDONLY(inode) && | |
244 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
245 | return -EROFS; | |
246 | ||
247 | /* | |
248 | * Nobody gets write access to an immutable file. | |
249 | */ | |
250 | if (IS_IMMUTABLE(inode)) | |
251 | return -EACCES; | |
252 | } | |
253 | ||
22590e41 MS |
254 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) { |
255 | /* | |
256 | * MAY_EXEC on regular files is denied if the fs is mounted | |
257 | * with the "noexec" flag. | |
258 | */ | |
c7eb2667 | 259 | if (mnt && (mnt->mnt_flags & MNT_NOEXEC)) |
22590e41 MS |
260 | return -EACCES; |
261 | } | |
a343bb77 | 262 | |
1da177e4 LT |
263 | /* Ordinary permission routines do not understand MAY_APPEND. */ |
264 | submask = mask & ~MAY_APPEND; | |
22590e41 | 265 | if (inode->i_op && inode->i_op->permission) { |
1da177e4 | 266 | retval = inode->i_op->permission(inode, submask, nd); |
22590e41 MS |
267 | if (!retval) { |
268 | /* | |
269 | * Exec permission on a regular file is denied if none | |
270 | * of the execute bits are set. | |
271 | * | |
272 | * This check should be done by the ->permission() | |
273 | * method. | |
274 | */ | |
275 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode) && | |
276 | !(inode->i_mode & S_IXUGO)) | |
277 | return -EACCES; | |
278 | } | |
279 | } else { | |
1da177e4 | 280 | retval = generic_permission(inode, submask, NULL); |
22590e41 | 281 | } |
1da177e4 LT |
282 | if (retval) |
283 | return retval; | |
284 | ||
08ce5f16 SH |
285 | retval = devcgroup_inode_permission(inode, mask); |
286 | if (retval) | |
287 | return retval; | |
288 | ||
1da177e4 LT |
289 | return security_inode_permission(inode, mask, nd); |
290 | } | |
291 | ||
e4543edd CH |
292 | /** |
293 | * vfs_permission - check for access rights to a given path | |
294 | * @nd: lookup result that describes the path | |
295 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
296 | * | |
297 | * Used to check for read/write/execute permissions on a path. | |
298 | * We use "fsuid" for this, letting us set arbitrary permissions | |
299 | * for filesystem access without changing the "normal" uids which | |
300 | * are used for other things. | |
301 | */ | |
302 | int vfs_permission(struct nameidata *nd, int mask) | |
303 | { | |
4ac91378 | 304 | return permission(nd->path.dentry->d_inode, mask, nd); |
e4543edd CH |
305 | } |
306 | ||
8c744fb8 CH |
307 | /** |
308 | * file_permission - check for additional access rights to a given file | |
309 | * @file: file to check access rights for | |
310 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
311 | * | |
312 | * Used to check for read/write/execute permissions on an already opened | |
313 | * file. | |
314 | * | |
315 | * Note: | |
316 | * Do not use this function in new code. All access checks should | |
317 | * be done using vfs_permission(). | |
318 | */ | |
319 | int file_permission(struct file *file, int mask) | |
320 | { | |
0f7fc9e4 | 321 | return permission(file->f_path.dentry->d_inode, mask, NULL); |
8c744fb8 CH |
322 | } |
323 | ||
1da177e4 LT |
324 | /* |
325 | * get_write_access() gets write permission for a file. | |
326 | * put_write_access() releases this write permission. | |
327 | * This is used for regular files. | |
328 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
329 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
330 | * can have the following values: | |
331 | * 0: no writers, no VM_DENYWRITE mappings | |
332 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
333 | * > 0: (i_writecount) users are writing to the file. | |
334 | * | |
335 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
336 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
337 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
338 | * to do the change if sign is wrong. Exclusion between them is provided by | |
339 | * the inode->i_lock spinlock. | |
340 | */ | |
341 | ||
342 | int get_write_access(struct inode * inode) | |
343 | { | |
344 | spin_lock(&inode->i_lock); | |
345 | if (atomic_read(&inode->i_writecount) < 0) { | |
346 | spin_unlock(&inode->i_lock); | |
347 | return -ETXTBSY; | |
348 | } | |
349 | atomic_inc(&inode->i_writecount); | |
350 | spin_unlock(&inode->i_lock); | |
351 | ||
352 | return 0; | |
353 | } | |
354 | ||
355 | int deny_write_access(struct file * file) | |
356 | { | |
0f7fc9e4 | 357 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
358 | |
359 | spin_lock(&inode->i_lock); | |
360 | if (atomic_read(&inode->i_writecount) > 0) { | |
361 | spin_unlock(&inode->i_lock); | |
362 | return -ETXTBSY; | |
363 | } | |
364 | atomic_dec(&inode->i_writecount); | |
365 | spin_unlock(&inode->i_lock); | |
366 | ||
367 | return 0; | |
368 | } | |
369 | ||
5dd784d0 JB |
370 | /** |
371 | * path_get - get a reference to a path | |
372 | * @path: path to get the reference to | |
373 | * | |
374 | * Given a path increment the reference count to the dentry and the vfsmount. | |
375 | */ | |
376 | void path_get(struct path *path) | |
377 | { | |
378 | mntget(path->mnt); | |
379 | dget(path->dentry); | |
380 | } | |
381 | EXPORT_SYMBOL(path_get); | |
382 | ||
1d957f9b JB |
383 | /** |
384 | * path_put - put a reference to a path | |
385 | * @path: path to put the reference to | |
386 | * | |
387 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
388 | */ | |
389 | void path_put(struct path *path) | |
1da177e4 | 390 | { |
1d957f9b JB |
391 | dput(path->dentry); |
392 | mntput(path->mnt); | |
1da177e4 | 393 | } |
1d957f9b | 394 | EXPORT_SYMBOL(path_put); |
1da177e4 | 395 | |
834f2a4a TM |
396 | /** |
397 | * release_open_intent - free up open intent resources | |
398 | * @nd: pointer to nameidata | |
399 | */ | |
400 | void release_open_intent(struct nameidata *nd) | |
401 | { | |
0f7fc9e4 | 402 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
403 | put_filp(nd->intent.open.file); |
404 | else | |
405 | fput(nd->intent.open.file); | |
406 | } | |
407 | ||
bcdc5e01 IK |
408 | static inline struct dentry * |
409 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
410 | { | |
411 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
412 | if (unlikely(status <= 0)) { | |
413 | /* | |
414 | * The dentry failed validation. | |
415 | * If d_revalidate returned 0 attempt to invalidate | |
416 | * the dentry otherwise d_revalidate is asking us | |
417 | * to return a fail status. | |
418 | */ | |
419 | if (!status) { | |
420 | if (!d_invalidate(dentry)) { | |
421 | dput(dentry); | |
422 | dentry = NULL; | |
423 | } | |
424 | } else { | |
425 | dput(dentry); | |
426 | dentry = ERR_PTR(status); | |
427 | } | |
428 | } | |
429 | return dentry; | |
430 | } | |
431 | ||
1da177e4 LT |
432 | /* |
433 | * Internal lookup() using the new generic dcache. | |
434 | * SMP-safe | |
435 | */ | |
436 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
437 | { | |
438 | struct dentry * dentry = __d_lookup(parent, name); | |
439 | ||
440 | /* lockess __d_lookup may fail due to concurrent d_move() | |
441 | * in some unrelated directory, so try with d_lookup | |
442 | */ | |
443 | if (!dentry) | |
444 | dentry = d_lookup(parent, name); | |
445 | ||
bcdc5e01 IK |
446 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
447 | dentry = do_revalidate(dentry, nd); | |
448 | ||
1da177e4 LT |
449 | return dentry; |
450 | } | |
451 | ||
452 | /* | |
453 | * Short-cut version of permission(), for calling by | |
454 | * path_walk(), when dcache lock is held. Combines parts | |
455 | * of permission() and generic_permission(), and tests ONLY for | |
456 | * MAY_EXEC permission. | |
457 | * | |
458 | * If appropriate, check DAC only. If not appropriate, or | |
459 | * short-cut DAC fails, then call permission() to do more | |
460 | * complete permission check. | |
461 | */ | |
858119e1 | 462 | static int exec_permission_lite(struct inode *inode, |
1da177e4 LT |
463 | struct nameidata *nd) |
464 | { | |
465 | umode_t mode = inode->i_mode; | |
466 | ||
467 | if (inode->i_op && inode->i_op->permission) | |
468 | return -EAGAIN; | |
469 | ||
470 | if (current->fsuid == inode->i_uid) | |
471 | mode >>= 6; | |
472 | else if (in_group_p(inode->i_gid)) | |
473 | mode >>= 3; | |
474 | ||
475 | if (mode & MAY_EXEC) | |
476 | goto ok; | |
477 | ||
478 | if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE)) | |
479 | goto ok; | |
480 | ||
481 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE)) | |
482 | goto ok; | |
483 | ||
484 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH)) | |
485 | goto ok; | |
486 | ||
487 | return -EACCES; | |
488 | ok: | |
489 | return security_inode_permission(inode, MAY_EXEC, nd); | |
490 | } | |
491 | ||
492 | /* | |
493 | * This is called when everything else fails, and we actually have | |
494 | * to go to the low-level filesystem to find out what we should do.. | |
495 | * | |
496 | * We get the directory semaphore, and after getting that we also | |
497 | * make sure that nobody added the entry to the dcache in the meantime.. | |
498 | * SMP-safe | |
499 | */ | |
500 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
501 | { | |
502 | struct dentry * result; | |
503 | struct inode *dir = parent->d_inode; | |
504 | ||
1b1dcc1b | 505 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
506 | /* |
507 | * First re-do the cached lookup just in case it was created | |
508 | * while we waited for the directory semaphore.. | |
509 | * | |
510 | * FIXME! This could use version numbering or similar to | |
511 | * avoid unnecessary cache lookups. | |
512 | * | |
513 | * The "dcache_lock" is purely to protect the RCU list walker | |
514 | * from concurrent renames at this point (we mustn't get false | |
515 | * negatives from the RCU list walk here, unlike the optimistic | |
516 | * fast walk). | |
517 | * | |
518 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
519 | */ | |
520 | result = d_lookup(parent, name); | |
521 | if (!result) { | |
522 | struct dentry * dentry = d_alloc(parent, name); | |
523 | result = ERR_PTR(-ENOMEM); | |
524 | if (dentry) { | |
525 | result = dir->i_op->lookup(dir, dentry, nd); | |
526 | if (result) | |
527 | dput(dentry); | |
528 | else | |
529 | result = dentry; | |
530 | } | |
1b1dcc1b | 531 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
532 | return result; |
533 | } | |
534 | ||
535 | /* | |
536 | * Uhhuh! Nasty case: the cache was re-populated while | |
537 | * we waited on the semaphore. Need to revalidate. | |
538 | */ | |
1b1dcc1b | 539 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 540 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
541 | result = do_revalidate(result, nd); |
542 | if (!result) | |
1da177e4 | 543 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
544 | } |
545 | return result; | |
546 | } | |
547 | ||
548 | static int __emul_lookup_dentry(const char *, struct nameidata *); | |
549 | ||
550 | /* SMP-safe */ | |
f1662356 | 551 | static __always_inline int |
1da177e4 LT |
552 | walk_init_root(const char *name, struct nameidata *nd) |
553 | { | |
e518ddb7 AM |
554 | struct fs_struct *fs = current->fs; |
555 | ||
556 | read_lock(&fs->lock); | |
6ac08c39 JB |
557 | if (fs->altroot.dentry && !(nd->flags & LOOKUP_NOALT)) { |
558 | nd->path = fs->altroot; | |
559 | path_get(&fs->altroot); | |
e518ddb7 | 560 | read_unlock(&fs->lock); |
1da177e4 LT |
561 | if (__emul_lookup_dentry(name,nd)) |
562 | return 0; | |
e518ddb7 | 563 | read_lock(&fs->lock); |
1da177e4 | 564 | } |
6ac08c39 JB |
565 | nd->path = fs->root; |
566 | path_get(&fs->root); | |
e518ddb7 | 567 | read_unlock(&fs->lock); |
1da177e4 LT |
568 | return 1; |
569 | } | |
570 | ||
a02f76c3 AV |
571 | /* |
572 | * Wrapper to retry pathname resolution whenever the underlying | |
573 | * file system returns an ESTALE. | |
574 | * | |
575 | * Retry the whole path once, forcing real lookup requests | |
576 | * instead of relying on the dcache. | |
577 | */ | |
578 | static __always_inline int link_path_walk(const char *name, struct nameidata *nd) | |
579 | { | |
580 | struct path save = nd->path; | |
581 | int result; | |
582 | ||
583 | /* make sure the stuff we saved doesn't go away */ | |
584 | dget(save.dentry); | |
585 | mntget(save.mnt); | |
586 | ||
587 | result = __link_path_walk(name, nd); | |
588 | if (result == -ESTALE) { | |
589 | /* nd->path had been dropped */ | |
590 | nd->path = save; | |
591 | dget(nd->path.dentry); | |
592 | mntget(nd->path.mnt); | |
593 | nd->flags |= LOOKUP_REVAL; | |
594 | result = __link_path_walk(name, nd); | |
595 | } | |
596 | ||
597 | path_put(&save); | |
598 | ||
599 | return result; | |
600 | } | |
601 | ||
f1662356 | 602 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
603 | { |
604 | int res = 0; | |
605 | char *name; | |
606 | if (IS_ERR(link)) | |
607 | goto fail; | |
608 | ||
609 | if (*link == '/') { | |
1d957f9b | 610 | path_put(&nd->path); |
1da177e4 LT |
611 | if (!walk_init_root(link, nd)) |
612 | /* weird __emul_prefix() stuff did it */ | |
613 | goto out; | |
614 | } | |
615 | res = link_path_walk(link, nd); | |
616 | out: | |
617 | if (nd->depth || res || nd->last_type!=LAST_NORM) | |
618 | return res; | |
619 | /* | |
620 | * If it is an iterative symlinks resolution in open_namei() we | |
621 | * have to copy the last component. And all that crap because of | |
622 | * bloody create() on broken symlinks. Furrfu... | |
623 | */ | |
624 | name = __getname(); | |
625 | if (unlikely(!name)) { | |
1d957f9b | 626 | path_put(&nd->path); |
1da177e4 LT |
627 | return -ENOMEM; |
628 | } | |
629 | strcpy(name, nd->last.name); | |
630 | nd->last.name = name; | |
631 | return 0; | |
632 | fail: | |
1d957f9b | 633 | path_put(&nd->path); |
1da177e4 LT |
634 | return PTR_ERR(link); |
635 | } | |
636 | ||
1d957f9b | 637 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
638 | { |
639 | dput(path->dentry); | |
4ac91378 | 640 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
641 | mntput(path->mnt); |
642 | } | |
643 | ||
644 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
645 | { | |
4ac91378 JB |
646 | dput(nd->path.dentry); |
647 | if (nd->path.mnt != path->mnt) | |
648 | mntput(nd->path.mnt); | |
649 | nd->path.mnt = path->mnt; | |
650 | nd->path.dentry = path->dentry; | |
051d3812 IK |
651 | } |
652 | ||
f1662356 | 653 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
654 | { |
655 | int error; | |
cc314eef | 656 | void *cookie; |
cd4e91d3 | 657 | struct dentry *dentry = path->dentry; |
1da177e4 | 658 | |
d671a1cb | 659 | touch_atime(path->mnt, dentry); |
1da177e4 | 660 | nd_set_link(nd, NULL); |
cd4e91d3 | 661 | |
4ac91378 | 662 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
663 | path_to_nameidata(path, nd); |
664 | dget(dentry); | |
665 | } | |
666 | mntget(path->mnt); | |
cc314eef LT |
667 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
668 | error = PTR_ERR(cookie); | |
669 | if (!IS_ERR(cookie)) { | |
1da177e4 | 670 | char *s = nd_get_link(nd); |
cc314eef | 671 | error = 0; |
1da177e4 LT |
672 | if (s) |
673 | error = __vfs_follow_link(nd, s); | |
674 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 675 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 676 | } |
09da5916 | 677 | path_put(path); |
1da177e4 LT |
678 | |
679 | return error; | |
680 | } | |
681 | ||
682 | /* | |
683 | * This limits recursive symlink follows to 8, while | |
684 | * limiting consecutive symlinks to 40. | |
685 | * | |
686 | * Without that kind of total limit, nasty chains of consecutive | |
687 | * symlinks can cause almost arbitrarily long lookups. | |
688 | */ | |
90ebe565 | 689 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
690 | { |
691 | int err = -ELOOP; | |
692 | if (current->link_count >= MAX_NESTED_LINKS) | |
693 | goto loop; | |
694 | if (current->total_link_count >= 40) | |
695 | goto loop; | |
696 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
697 | cond_resched(); | |
90ebe565 | 698 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
699 | if (err) |
700 | goto loop; | |
701 | current->link_count++; | |
702 | current->total_link_count++; | |
703 | nd->depth++; | |
cd4e91d3 | 704 | err = __do_follow_link(path, nd); |
839d9f93 AV |
705 | current->link_count--; |
706 | nd->depth--; | |
1da177e4 LT |
707 | return err; |
708 | loop: | |
1d957f9b JB |
709 | path_put_conditional(path, nd); |
710 | path_put(&nd->path); | |
1da177e4 LT |
711 | return err; |
712 | } | |
713 | ||
714 | int follow_up(struct vfsmount **mnt, struct dentry **dentry) | |
715 | { | |
716 | struct vfsmount *parent; | |
717 | struct dentry *mountpoint; | |
718 | spin_lock(&vfsmount_lock); | |
719 | parent=(*mnt)->mnt_parent; | |
720 | if (parent == *mnt) { | |
721 | spin_unlock(&vfsmount_lock); | |
722 | return 0; | |
723 | } | |
724 | mntget(parent); | |
725 | mountpoint=dget((*mnt)->mnt_mountpoint); | |
726 | spin_unlock(&vfsmount_lock); | |
727 | dput(*dentry); | |
728 | *dentry = mountpoint; | |
729 | mntput(*mnt); | |
730 | *mnt = parent; | |
731 | return 1; | |
732 | } | |
733 | ||
734 | /* no need for dcache_lock, as serialization is taken care in | |
735 | * namespace.c | |
736 | */ | |
463ffb2e AV |
737 | static int __follow_mount(struct path *path) |
738 | { | |
739 | int res = 0; | |
740 | while (d_mountpoint(path->dentry)) { | |
741 | struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry); | |
742 | if (!mounted) | |
743 | break; | |
744 | dput(path->dentry); | |
745 | if (res) | |
746 | mntput(path->mnt); | |
747 | path->mnt = mounted; | |
748 | path->dentry = dget(mounted->mnt_root); | |
749 | res = 1; | |
750 | } | |
751 | return res; | |
752 | } | |
753 | ||
58c465eb | 754 | static void follow_mount(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 | 755 | { |
1da177e4 LT |
756 | while (d_mountpoint(*dentry)) { |
757 | struct vfsmount *mounted = lookup_mnt(*mnt, *dentry); | |
758 | if (!mounted) | |
759 | break; | |
58c465eb | 760 | dput(*dentry); |
1da177e4 LT |
761 | mntput(*mnt); |
762 | *mnt = mounted; | |
1da177e4 | 763 | *dentry = dget(mounted->mnt_root); |
1da177e4 | 764 | } |
1da177e4 LT |
765 | } |
766 | ||
767 | /* no need for dcache_lock, as serialization is taken care in | |
768 | * namespace.c | |
769 | */ | |
e13b210f | 770 | int follow_down(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 LT |
771 | { |
772 | struct vfsmount *mounted; | |
773 | ||
774 | mounted = lookup_mnt(*mnt, *dentry); | |
775 | if (mounted) { | |
e13b210f | 776 | dput(*dentry); |
1da177e4 LT |
777 | mntput(*mnt); |
778 | *mnt = mounted; | |
1da177e4 LT |
779 | *dentry = dget(mounted->mnt_root); |
780 | return 1; | |
781 | } | |
782 | return 0; | |
783 | } | |
784 | ||
f1662356 | 785 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 786 | { |
e518ddb7 AM |
787 | struct fs_struct *fs = current->fs; |
788 | ||
1da177e4 LT |
789 | while(1) { |
790 | struct vfsmount *parent; | |
4ac91378 | 791 | struct dentry *old = nd->path.dentry; |
1da177e4 | 792 | |
e518ddb7 | 793 | read_lock(&fs->lock); |
6ac08c39 JB |
794 | if (nd->path.dentry == fs->root.dentry && |
795 | nd->path.mnt == fs->root.mnt) { | |
e518ddb7 | 796 | read_unlock(&fs->lock); |
1da177e4 LT |
797 | break; |
798 | } | |
e518ddb7 | 799 | read_unlock(&fs->lock); |
1da177e4 | 800 | spin_lock(&dcache_lock); |
4ac91378 JB |
801 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
802 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
803 | spin_unlock(&dcache_lock); |
804 | dput(old); | |
805 | break; | |
806 | } | |
807 | spin_unlock(&dcache_lock); | |
808 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
809 | parent = nd->path.mnt->mnt_parent; |
810 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
811 | spin_unlock(&vfsmount_lock); |
812 | break; | |
813 | } | |
814 | mntget(parent); | |
4ac91378 | 815 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
816 | spin_unlock(&vfsmount_lock); |
817 | dput(old); | |
4ac91378 JB |
818 | mntput(nd->path.mnt); |
819 | nd->path.mnt = parent; | |
1da177e4 | 820 | } |
4ac91378 | 821 | follow_mount(&nd->path.mnt, &nd->path.dentry); |
1da177e4 LT |
822 | } |
823 | ||
1da177e4 LT |
824 | /* |
825 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
826 | * small and for now I'd prefer to have fast path as straight as possible. | |
827 | * It _is_ time-critical. | |
828 | */ | |
829 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
830 | struct path *path) | |
831 | { | |
4ac91378 JB |
832 | struct vfsmount *mnt = nd->path.mnt; |
833 | struct dentry *dentry = __d_lookup(nd->path.dentry, name); | |
1da177e4 LT |
834 | |
835 | if (!dentry) | |
836 | goto need_lookup; | |
837 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
838 | goto need_revalidate; | |
839 | done: | |
840 | path->mnt = mnt; | |
841 | path->dentry = dentry; | |
634ee701 | 842 | __follow_mount(path); |
1da177e4 LT |
843 | return 0; |
844 | ||
845 | need_lookup: | |
4ac91378 | 846 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
847 | if (IS_ERR(dentry)) |
848 | goto fail; | |
849 | goto done; | |
850 | ||
851 | need_revalidate: | |
bcdc5e01 IK |
852 | dentry = do_revalidate(dentry, nd); |
853 | if (!dentry) | |
854 | goto need_lookup; | |
855 | if (IS_ERR(dentry)) | |
856 | goto fail; | |
857 | goto done; | |
1da177e4 LT |
858 | |
859 | fail: | |
860 | return PTR_ERR(dentry); | |
861 | } | |
862 | ||
863 | /* | |
864 | * Name resolution. | |
ea3834d9 PM |
865 | * This is the basic name resolution function, turning a pathname into |
866 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 867 | * |
ea3834d9 PM |
868 | * Returns 0 and nd will have valid dentry and mnt on success. |
869 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 870 | */ |
fc9b52cd | 871 | static int __link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
872 | { |
873 | struct path next; | |
874 | struct inode *inode; | |
875 | int err; | |
876 | unsigned int lookup_flags = nd->flags; | |
877 | ||
878 | while (*name=='/') | |
879 | name++; | |
880 | if (!*name) | |
881 | goto return_reval; | |
882 | ||
4ac91378 | 883 | inode = nd->path.dentry->d_inode; |
1da177e4 | 884 | if (nd->depth) |
f55eab82 | 885 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
886 | |
887 | /* At this point we know we have a real path component. */ | |
888 | for(;;) { | |
889 | unsigned long hash; | |
890 | struct qstr this; | |
891 | unsigned int c; | |
892 | ||
cdce5d6b | 893 | nd->flags |= LOOKUP_CONTINUE; |
1da177e4 | 894 | err = exec_permission_lite(inode, nd); |
e4543edd CH |
895 | if (err == -EAGAIN) |
896 | err = vfs_permission(nd, MAY_EXEC); | |
1da177e4 LT |
897 | if (err) |
898 | break; | |
899 | ||
900 | this.name = name; | |
901 | c = *(const unsigned char *)name; | |
902 | ||
903 | hash = init_name_hash(); | |
904 | do { | |
905 | name++; | |
906 | hash = partial_name_hash(c, hash); | |
907 | c = *(const unsigned char *)name; | |
908 | } while (c && (c != '/')); | |
909 | this.len = name - (const char *) this.name; | |
910 | this.hash = end_name_hash(hash); | |
911 | ||
912 | /* remove trailing slashes? */ | |
913 | if (!c) | |
914 | goto last_component; | |
915 | while (*++name == '/'); | |
916 | if (!*name) | |
917 | goto last_with_slashes; | |
918 | ||
919 | /* | |
920 | * "." and ".." are special - ".." especially so because it has | |
921 | * to be able to know about the current root directory and | |
922 | * parent relationships. | |
923 | */ | |
924 | if (this.name[0] == '.') switch (this.len) { | |
925 | default: | |
926 | break; | |
927 | case 2: | |
928 | if (this.name[1] != '.') | |
929 | break; | |
58c465eb | 930 | follow_dotdot(nd); |
4ac91378 | 931 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
932 | /* fallthrough */ |
933 | case 1: | |
934 | continue; | |
935 | } | |
936 | /* | |
937 | * See if the low-level filesystem might want | |
938 | * to use its own hash.. | |
939 | */ | |
4ac91378 JB |
940 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
941 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
942 | &this); | |
1da177e4 LT |
943 | if (err < 0) |
944 | break; | |
945 | } | |
1da177e4 LT |
946 | /* This does the actual lookups.. */ |
947 | err = do_lookup(nd, &this, &next); | |
948 | if (err) | |
949 | break; | |
1da177e4 LT |
950 | |
951 | err = -ENOENT; | |
952 | inode = next.dentry->d_inode; | |
953 | if (!inode) | |
954 | goto out_dput; | |
955 | err = -ENOTDIR; | |
956 | if (!inode->i_op) | |
957 | goto out_dput; | |
958 | ||
959 | if (inode->i_op->follow_link) { | |
90ebe565 | 960 | err = do_follow_link(&next, nd); |
1da177e4 LT |
961 | if (err) |
962 | goto return_err; | |
963 | err = -ENOENT; | |
4ac91378 | 964 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
965 | if (!inode) |
966 | break; | |
967 | err = -ENOTDIR; | |
968 | if (!inode->i_op) | |
969 | break; | |
09dd17d3 MS |
970 | } else |
971 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
972 | err = -ENOTDIR; |
973 | if (!inode->i_op->lookup) | |
974 | break; | |
975 | continue; | |
976 | /* here ends the main loop */ | |
977 | ||
978 | last_with_slashes: | |
979 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
980 | last_component: | |
f55eab82 TM |
981 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
982 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
983 | if (lookup_flags & LOOKUP_PARENT) |
984 | goto lookup_parent; | |
985 | if (this.name[0] == '.') switch (this.len) { | |
986 | default: | |
987 | break; | |
988 | case 2: | |
989 | if (this.name[1] != '.') | |
990 | break; | |
58c465eb | 991 | follow_dotdot(nd); |
4ac91378 | 992 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
993 | /* fallthrough */ |
994 | case 1: | |
995 | goto return_reval; | |
996 | } | |
4ac91378 JB |
997 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
998 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
999 | &this); | |
1da177e4 LT |
1000 | if (err < 0) |
1001 | break; | |
1002 | } | |
1003 | err = do_lookup(nd, &this, &next); | |
1004 | if (err) | |
1005 | break; | |
1da177e4 LT |
1006 | inode = next.dentry->d_inode; |
1007 | if ((lookup_flags & LOOKUP_FOLLOW) | |
1008 | && inode && inode->i_op && inode->i_op->follow_link) { | |
90ebe565 | 1009 | err = do_follow_link(&next, nd); |
1da177e4 LT |
1010 | if (err) |
1011 | goto return_err; | |
4ac91378 | 1012 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
1013 | } else |
1014 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
1015 | err = -ENOENT; |
1016 | if (!inode) | |
1017 | break; | |
1018 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
1019 | err = -ENOTDIR; | |
1020 | if (!inode->i_op || !inode->i_op->lookup) | |
1021 | break; | |
1022 | } | |
1023 | goto return_base; | |
1024 | lookup_parent: | |
1025 | nd->last = this; | |
1026 | nd->last_type = LAST_NORM; | |
1027 | if (this.name[0] != '.') | |
1028 | goto return_base; | |
1029 | if (this.len == 1) | |
1030 | nd->last_type = LAST_DOT; | |
1031 | else if (this.len == 2 && this.name[1] == '.') | |
1032 | nd->last_type = LAST_DOTDOT; | |
1033 | else | |
1034 | goto return_base; | |
1035 | return_reval: | |
1036 | /* | |
1037 | * We bypassed the ordinary revalidation routines. | |
1038 | * We may need to check the cached dentry for staleness. | |
1039 | */ | |
4ac91378 JB |
1040 | if (nd->path.dentry && nd->path.dentry->d_sb && |
1041 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
1042 | err = -ESTALE; |
1043 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
1044 | if (!nd->path.dentry->d_op->d_revalidate( |
1045 | nd->path.dentry, nd)) | |
1da177e4 LT |
1046 | break; |
1047 | } | |
1048 | return_base: | |
1049 | return 0; | |
1050 | out_dput: | |
1d957f9b | 1051 | path_put_conditional(&next, nd); |
1da177e4 LT |
1052 | break; |
1053 | } | |
1d957f9b | 1054 | path_put(&nd->path); |
1da177e4 LT |
1055 | return_err: |
1056 | return err; | |
1057 | } | |
1058 | ||
fc9b52cd | 1059 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1060 | { |
1061 | current->total_link_count = 0; | |
1062 | return link_path_walk(name, nd); | |
1063 | } | |
1064 | ||
ea3834d9 PM |
1065 | /* |
1066 | * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if | |
1067 | * everything is done. Returns 0 and drops input nd, if lookup failed; | |
1068 | */ | |
1da177e4 LT |
1069 | static int __emul_lookup_dentry(const char *name, struct nameidata *nd) |
1070 | { | |
1071 | if (path_walk(name, nd)) | |
1072 | return 0; /* something went wrong... */ | |
1073 | ||
4ac91378 JB |
1074 | if (!nd->path.dentry->d_inode || |
1075 | S_ISDIR(nd->path.dentry->d_inode->i_mode)) { | |
09da5916 | 1076 | struct path old_path = nd->path; |
1da177e4 LT |
1077 | struct qstr last = nd->last; |
1078 | int last_type = nd->last_type; | |
e518ddb7 AM |
1079 | struct fs_struct *fs = current->fs; |
1080 | ||
1da177e4 | 1081 | /* |
e518ddb7 AM |
1082 | * NAME was not found in alternate root or it's a directory. |
1083 | * Try to find it in the normal root: | |
1da177e4 LT |
1084 | */ |
1085 | nd->last_type = LAST_ROOT; | |
e518ddb7 | 1086 | read_lock(&fs->lock); |
6ac08c39 JB |
1087 | nd->path = fs->root; |
1088 | path_get(&fs->root); | |
e518ddb7 | 1089 | read_unlock(&fs->lock); |
1da177e4 | 1090 | if (path_walk(name, nd) == 0) { |
4ac91378 | 1091 | if (nd->path.dentry->d_inode) { |
09da5916 | 1092 | path_put(&old_path); |
1da177e4 LT |
1093 | return 1; |
1094 | } | |
1d957f9b | 1095 | path_put(&nd->path); |
1da177e4 | 1096 | } |
09da5916 | 1097 | nd->path = old_path; |
1da177e4 LT |
1098 | nd->last = last; |
1099 | nd->last_type = last_type; | |
1100 | } | |
1101 | return 1; | |
1102 | } | |
1103 | ||
1104 | void set_fs_altroot(void) | |
1105 | { | |
1106 | char *emul = __emul_prefix(); | |
1107 | struct nameidata nd; | |
6ac08c39 | 1108 | struct path path = {}, old_path; |
1da177e4 | 1109 | int err; |
e518ddb7 | 1110 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1111 | |
1112 | if (!emul) | |
1113 | goto set_it; | |
1114 | err = path_lookup(emul, LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_NOALT, &nd); | |
6ac08c39 JB |
1115 | if (!err) |
1116 | path = nd.path; | |
1da177e4 | 1117 | set_it: |
e518ddb7 | 1118 | write_lock(&fs->lock); |
6ac08c39 JB |
1119 | old_path = fs->altroot; |
1120 | fs->altroot = path; | |
e518ddb7 | 1121 | write_unlock(&fs->lock); |
6ac08c39 JB |
1122 | if (old_path.dentry) |
1123 | path_put(&old_path); | |
1da177e4 LT |
1124 | } |
1125 | ||
ea3834d9 | 1126 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
fc9b52cd | 1127 | static int do_path_lookup(int dfd, const char *name, |
5590ff0d | 1128 | unsigned int flags, struct nameidata *nd) |
1da177e4 | 1129 | { |
ea3834d9 | 1130 | int retval = 0; |
170aa3d0 UD |
1131 | int fput_needed; |
1132 | struct file *file; | |
e518ddb7 | 1133 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1134 | |
1135 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1136 | nd->flags = flags; | |
1137 | nd->depth = 0; | |
1138 | ||
1da177e4 | 1139 | if (*name=='/') { |
e518ddb7 | 1140 | read_lock(&fs->lock); |
6ac08c39 JB |
1141 | if (fs->altroot.dentry && !(nd->flags & LOOKUP_NOALT)) { |
1142 | nd->path = fs->altroot; | |
1143 | path_get(&fs->altroot); | |
e518ddb7 | 1144 | read_unlock(&fs->lock); |
1da177e4 | 1145 | if (__emul_lookup_dentry(name,nd)) |
ea3834d9 | 1146 | goto out; /* found in altroot */ |
e518ddb7 | 1147 | read_lock(&fs->lock); |
1da177e4 | 1148 | } |
6ac08c39 JB |
1149 | nd->path = fs->root; |
1150 | path_get(&fs->root); | |
e518ddb7 | 1151 | read_unlock(&fs->lock); |
5590ff0d | 1152 | } else if (dfd == AT_FDCWD) { |
e518ddb7 | 1153 | read_lock(&fs->lock); |
6ac08c39 JB |
1154 | nd->path = fs->pwd; |
1155 | path_get(&fs->pwd); | |
e518ddb7 | 1156 | read_unlock(&fs->lock); |
5590ff0d | 1157 | } else { |
5590ff0d UD |
1158 | struct dentry *dentry; |
1159 | ||
1160 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1161 | retval = -EBADF; |
1162 | if (!file) | |
6d09bb62 | 1163 | goto out_fail; |
5590ff0d | 1164 | |
0f7fc9e4 | 1165 | dentry = file->f_path.dentry; |
5590ff0d | 1166 | |
170aa3d0 UD |
1167 | retval = -ENOTDIR; |
1168 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1169 | goto fput_fail; |
5590ff0d UD |
1170 | |
1171 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1172 | if (retval) |
6d09bb62 | 1173 | goto fput_fail; |
5590ff0d | 1174 | |
5dd784d0 JB |
1175 | nd->path = file->f_path; |
1176 | path_get(&file->f_path); | |
5590ff0d UD |
1177 | |
1178 | fput_light(file, fput_needed); | |
1da177e4 | 1179 | } |
2dfdd266 JJS |
1180 | |
1181 | retval = path_walk(name, nd); | |
ea3834d9 | 1182 | out: |
4ac91378 JB |
1183 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1184 | nd->path.dentry->d_inode)) | |
1185 | audit_inode(name, nd->path.dentry); | |
6d09bb62 | 1186 | out_fail: |
170aa3d0 UD |
1187 | return retval; |
1188 | ||
6d09bb62 | 1189 | fput_fail: |
170aa3d0 | 1190 | fput_light(file, fput_needed); |
6d09bb62 | 1191 | goto out_fail; |
1da177e4 LT |
1192 | } |
1193 | ||
fc9b52cd | 1194 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1195 | struct nameidata *nd) |
1196 | { | |
1197 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1198 | } | |
1199 | ||
16f18200 JJS |
1200 | /** |
1201 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1202 | * @dentry: pointer to dentry of the base directory | |
1203 | * @mnt: pointer to vfs mount of the base directory | |
1204 | * @name: pointer to file name | |
1205 | * @flags: lookup flags | |
1206 | * @nd: pointer to nameidata | |
1207 | */ | |
1208 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1209 | const char *name, unsigned int flags, | |
1210 | struct nameidata *nd) | |
1211 | { | |
1212 | int retval; | |
1213 | ||
1214 | /* same as do_path_lookup */ | |
1215 | nd->last_type = LAST_ROOT; | |
1216 | nd->flags = flags; | |
1217 | nd->depth = 0; | |
1218 | ||
4ac91378 JB |
1219 | nd->path.mnt = mntget(mnt); |
1220 | nd->path.dentry = dget(dentry); | |
16f18200 JJS |
1221 | |
1222 | retval = path_walk(name, nd); | |
4ac91378 JB |
1223 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1224 | nd->path.dentry->d_inode)) | |
1225 | audit_inode(name, nd->path.dentry); | |
16f18200 JJS |
1226 | |
1227 | return retval; | |
1228 | ||
1229 | } | |
1230 | ||
5590ff0d UD |
1231 | static int __path_lookup_intent_open(int dfd, const char *name, |
1232 | unsigned int lookup_flags, struct nameidata *nd, | |
1233 | int open_flags, int create_mode) | |
834f2a4a TM |
1234 | { |
1235 | struct file *filp = get_empty_filp(); | |
1236 | int err; | |
1237 | ||
1238 | if (filp == NULL) | |
1239 | return -ENFILE; | |
1240 | nd->intent.open.file = filp; | |
1241 | nd->intent.open.flags = open_flags; | |
1242 | nd->intent.open.create_mode = create_mode; | |
5590ff0d | 1243 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1244 | if (IS_ERR(nd->intent.open.file)) { |
1245 | if (err == 0) { | |
1246 | err = PTR_ERR(nd->intent.open.file); | |
1d957f9b | 1247 | path_put(&nd->path); |
834f2a4a TM |
1248 | } |
1249 | } else if (err != 0) | |
1250 | release_open_intent(nd); | |
1251 | return err; | |
1252 | } | |
1253 | ||
1254 | /** | |
1255 | * path_lookup_open - lookup a file path with open intent | |
7045f37b | 1256 | * @dfd: the directory to use as base, or AT_FDCWD |
834f2a4a TM |
1257 | * @name: pointer to file name |
1258 | * @lookup_flags: lookup intent flags | |
1259 | * @nd: pointer to nameidata | |
1260 | * @open_flags: open intent flags | |
1261 | */ | |
5590ff0d | 1262 | int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags, |
834f2a4a TM |
1263 | struct nameidata *nd, int open_flags) |
1264 | { | |
5590ff0d | 1265 | return __path_lookup_intent_open(dfd, name, lookup_flags, nd, |
834f2a4a TM |
1266 | open_flags, 0); |
1267 | } | |
1268 | ||
1269 | /** | |
1270 | * path_lookup_create - lookup a file path with open + create intent | |
7045f37b | 1271 | * @dfd: the directory to use as base, or AT_FDCWD |
834f2a4a TM |
1272 | * @name: pointer to file name |
1273 | * @lookup_flags: lookup intent flags | |
1274 | * @nd: pointer to nameidata | |
1275 | * @open_flags: open intent flags | |
1276 | * @create_mode: create intent flags | |
1277 | */ | |
5590ff0d UD |
1278 | static int path_lookup_create(int dfd, const char *name, |
1279 | unsigned int lookup_flags, struct nameidata *nd, | |
1280 | int open_flags, int create_mode) | |
834f2a4a | 1281 | { |
5590ff0d UD |
1282 | return __path_lookup_intent_open(dfd, name, lookup_flags|LOOKUP_CREATE, |
1283 | nd, open_flags, create_mode); | |
834f2a4a TM |
1284 | } |
1285 | ||
1286 | int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags, | |
1287 | struct nameidata *nd, int open_flags) | |
1288 | { | |
1289 | char *tmp = getname(name); | |
1290 | int err = PTR_ERR(tmp); | |
1291 | ||
1292 | if (!IS_ERR(tmp)) { | |
5590ff0d | 1293 | err = __path_lookup_intent_open(AT_FDCWD, tmp, lookup_flags, nd, open_flags, 0); |
834f2a4a TM |
1294 | putname(tmp); |
1295 | } | |
1296 | return err; | |
1297 | } | |
1298 | ||
eead1911 CH |
1299 | static struct dentry *__lookup_hash(struct qstr *name, |
1300 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1301 | { |
057f6c01 | 1302 | struct dentry *dentry; |
1da177e4 LT |
1303 | struct inode *inode; |
1304 | int err; | |
1305 | ||
1306 | inode = base->d_inode; | |
1da177e4 LT |
1307 | |
1308 | /* | |
1309 | * See if the low-level filesystem might want | |
1310 | * to use its own hash.. | |
1311 | */ | |
1312 | if (base->d_op && base->d_op->d_hash) { | |
1313 | err = base->d_op->d_hash(base, name); | |
1314 | dentry = ERR_PTR(err); | |
1315 | if (err < 0) | |
1316 | goto out; | |
1317 | } | |
1318 | ||
1319 | dentry = cached_lookup(base, name, nd); | |
1320 | if (!dentry) { | |
1321 | struct dentry *new = d_alloc(base, name); | |
1322 | dentry = ERR_PTR(-ENOMEM); | |
1323 | if (!new) | |
1324 | goto out; | |
1325 | dentry = inode->i_op->lookup(inode, new, nd); | |
1326 | if (!dentry) | |
1327 | dentry = new; | |
1328 | else | |
1329 | dput(new); | |
1330 | } | |
1331 | out: | |
1332 | return dentry; | |
1333 | } | |
1334 | ||
057f6c01 JM |
1335 | /* |
1336 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1337 | * needs parent already locked. Doesn't follow mounts. | |
1338 | * SMP-safe. | |
1339 | */ | |
eead1911 | 1340 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1341 | { |
057f6c01 JM |
1342 | int err; |
1343 | ||
4ac91378 | 1344 | err = permission(nd->path.dentry->d_inode, MAY_EXEC, nd); |
057f6c01 | 1345 | if (err) |
eead1911 | 1346 | return ERR_PTR(err); |
4ac91378 | 1347 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1348 | } |
1349 | ||
eead1911 CH |
1350 | static int __lookup_one_len(const char *name, struct qstr *this, |
1351 | struct dentry *base, int len) | |
1da177e4 LT |
1352 | { |
1353 | unsigned long hash; | |
1da177e4 LT |
1354 | unsigned int c; |
1355 | ||
057f6c01 JM |
1356 | this->name = name; |
1357 | this->len = len; | |
1da177e4 | 1358 | if (!len) |
057f6c01 | 1359 | return -EACCES; |
1da177e4 LT |
1360 | |
1361 | hash = init_name_hash(); | |
1362 | while (len--) { | |
1363 | c = *(const unsigned char *)name++; | |
1364 | if (c == '/' || c == '\0') | |
057f6c01 | 1365 | return -EACCES; |
1da177e4 LT |
1366 | hash = partial_name_hash(c, hash); |
1367 | } | |
057f6c01 JM |
1368 | this->hash = end_name_hash(hash); |
1369 | return 0; | |
1370 | } | |
1da177e4 | 1371 | |
eead1911 | 1372 | /** |
a6b91919 | 1373 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1374 | * @name: pathname component to lookup |
1375 | * @base: base directory to lookup from | |
1376 | * @len: maximum length @len should be interpreted to | |
1377 | * | |
a6b91919 RD |
1378 | * Note that this routine is purely a helper for filesystem usage and should |
1379 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1380 | * nameidata argument is passed to the filesystem methods and a filesystem |
1381 | * using this helper needs to be prepared for that. | |
1382 | */ | |
057f6c01 JM |
1383 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1384 | { | |
1385 | int err; | |
1386 | struct qstr this; | |
1387 | ||
1388 | err = __lookup_one_len(name, &this, base, len); | |
eead1911 CH |
1389 | if (err) |
1390 | return ERR_PTR(err); | |
1391 | ||
1392 | err = permission(base->d_inode, MAY_EXEC, NULL); | |
057f6c01 JM |
1393 | if (err) |
1394 | return ERR_PTR(err); | |
49705b77 | 1395 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1396 | } |
1397 | ||
eead1911 CH |
1398 | /** |
1399 | * lookup_one_noperm - bad hack for sysfs | |
1400 | * @name: pathname component to lookup | |
1401 | * @base: base directory to lookup from | |
1402 | * | |
1403 | * This is a variant of lookup_one_len that doesn't perform any permission | |
1404 | * checks. It's a horrible hack to work around the braindead sysfs | |
1405 | * architecture and should not be used anywhere else. | |
1406 | * | |
1407 | * DON'T USE THIS FUNCTION EVER, thanks. | |
1408 | */ | |
1409 | struct dentry *lookup_one_noperm(const char *name, struct dentry *base) | |
057f6c01 JM |
1410 | { |
1411 | int err; | |
1412 | struct qstr this; | |
1413 | ||
eead1911 | 1414 | err = __lookup_one_len(name, &this, base, strlen(name)); |
057f6c01 JM |
1415 | if (err) |
1416 | return ERR_PTR(err); | |
eead1911 | 1417 | return __lookup_hash(&this, base, NULL); |
1da177e4 LT |
1418 | } |
1419 | ||
fc9b52cd | 1420 | int __user_walk_fd(int dfd, const char __user *name, unsigned flags, |
5590ff0d | 1421 | struct nameidata *nd) |
1da177e4 LT |
1422 | { |
1423 | char *tmp = getname(name); | |
1424 | int err = PTR_ERR(tmp); | |
1425 | ||
1426 | if (!IS_ERR(tmp)) { | |
5590ff0d | 1427 | err = do_path_lookup(dfd, tmp, flags, nd); |
1da177e4 LT |
1428 | putname(tmp); |
1429 | } | |
1430 | return err; | |
1431 | } | |
1432 | ||
fc9b52cd | 1433 | int __user_walk(const char __user *name, unsigned flags, struct nameidata *nd) |
5590ff0d UD |
1434 | { |
1435 | return __user_walk_fd(AT_FDCWD, name, flags, nd); | |
1436 | } | |
1437 | ||
1da177e4 LT |
1438 | /* |
1439 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1440 | * minimal. | |
1441 | */ | |
1442 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1443 | { | |
1444 | if (!(dir->i_mode & S_ISVTX)) | |
1445 | return 0; | |
1446 | if (inode->i_uid == current->fsuid) | |
1447 | return 0; | |
1448 | if (dir->i_uid == current->fsuid) | |
1449 | return 0; | |
1450 | return !capable(CAP_FOWNER); | |
1451 | } | |
1452 | ||
1453 | /* | |
1454 | * Check whether we can remove a link victim from directory dir, check | |
1455 | * whether the type of victim is right. | |
1456 | * 1. We can't do it if dir is read-only (done in permission()) | |
1457 | * 2. We should have write and exec permissions on dir | |
1458 | * 3. We can't remove anything from append-only dir | |
1459 | * 4. We can't do anything with immutable dir (done in permission()) | |
1460 | * 5. If the sticky bit on dir is set we should either | |
1461 | * a. be owner of dir, or | |
1462 | * b. be owner of victim, or | |
1463 | * c. have CAP_FOWNER capability | |
1464 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1465 | * links pointing to it. | |
1466 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1467 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1468 | * 9. We can't remove a root or mountpoint. | |
1469 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1470 | * nfs_async_unlink(). | |
1471 | */ | |
858119e1 | 1472 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1473 | { |
1474 | int error; | |
1475 | ||
1476 | if (!victim->d_inode) | |
1477 | return -ENOENT; | |
1478 | ||
1479 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1480 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 LT |
1481 | |
1482 | error = permission(dir,MAY_WRITE | MAY_EXEC, NULL); | |
1483 | if (error) | |
1484 | return error; | |
1485 | if (IS_APPEND(dir)) | |
1486 | return -EPERM; | |
1487 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
1488 | IS_IMMUTABLE(victim->d_inode)) | |
1489 | return -EPERM; | |
1490 | if (isdir) { | |
1491 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1492 | return -ENOTDIR; | |
1493 | if (IS_ROOT(victim)) | |
1494 | return -EBUSY; | |
1495 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1496 | return -EISDIR; | |
1497 | if (IS_DEADDIR(dir)) | |
1498 | return -ENOENT; | |
1499 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1500 | return -EBUSY; | |
1501 | return 0; | |
1502 | } | |
1503 | ||
1504 | /* Check whether we can create an object with dentry child in directory | |
1505 | * dir. | |
1506 | * 1. We can't do it if child already exists (open has special treatment for | |
1507 | * this case, but since we are inlined it's OK) | |
1508 | * 2. We can't do it if dir is read-only (done in permission()) | |
1509 | * 3. We should have write and exec permissions on dir | |
1510 | * 4. We can't do it if dir is immutable (done in permission()) | |
1511 | */ | |
1512 | static inline int may_create(struct inode *dir, struct dentry *child, | |
1513 | struct nameidata *nd) | |
1514 | { | |
1515 | if (child->d_inode) | |
1516 | return -EEXIST; | |
1517 | if (IS_DEADDIR(dir)) | |
1518 | return -ENOENT; | |
1519 | return permission(dir,MAY_WRITE | MAY_EXEC, nd); | |
1520 | } | |
1521 | ||
1522 | /* | |
1da177e4 LT |
1523 | * O_DIRECTORY translates into forcing a directory lookup. |
1524 | */ | |
1525 | static inline int lookup_flags(unsigned int f) | |
1526 | { | |
1527 | unsigned long retval = LOOKUP_FOLLOW; | |
1528 | ||
1529 | if (f & O_NOFOLLOW) | |
1530 | retval &= ~LOOKUP_FOLLOW; | |
1531 | ||
1da177e4 LT |
1532 | if (f & O_DIRECTORY) |
1533 | retval |= LOOKUP_DIRECTORY; | |
1534 | ||
1535 | return retval; | |
1536 | } | |
1537 | ||
1538 | /* | |
1539 | * p1 and p2 should be directories on the same fs. | |
1540 | */ | |
1541 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1542 | { | |
1543 | struct dentry *p; | |
1544 | ||
1545 | if (p1 == p2) { | |
f2eace23 | 1546 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1547 | return NULL; |
1548 | } | |
1549 | ||
a11f3a05 | 1550 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1551 | |
1552 | for (p = p1; p->d_parent != p; p = p->d_parent) { | |
1553 | if (p->d_parent == p2) { | |
f2eace23 IM |
1554 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); |
1555 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1556 | return p; |
1557 | } | |
1558 | } | |
1559 | ||
1560 | for (p = p2; p->d_parent != p; p = p->d_parent) { | |
1561 | if (p->d_parent == p1) { | |
f2eace23 IM |
1562 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1563 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1564 | return p; |
1565 | } | |
1566 | } | |
1567 | ||
f2eace23 IM |
1568 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1569 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1570 | return NULL; |
1571 | } | |
1572 | ||
1573 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1574 | { | |
1b1dcc1b | 1575 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1576 | if (p1 != p2) { |
1b1dcc1b | 1577 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1578 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1579 | } |
1580 | } | |
1581 | ||
1582 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1583 | struct nameidata *nd) | |
1584 | { | |
1585 | int error = may_create(dir, dentry, nd); | |
1586 | ||
1587 | if (error) | |
1588 | return error; | |
1589 | ||
1590 | if (!dir->i_op || !dir->i_op->create) | |
1591 | return -EACCES; /* shouldn't it be ENOSYS? */ | |
1592 | mode &= S_IALLUGO; | |
1593 | mode |= S_IFREG; | |
1594 | error = security_inode_create(dir, dentry, mode); | |
1595 | if (error) | |
1596 | return error; | |
1597 | DQUOT_INIT(dir); | |
1598 | error = dir->i_op->create(dir, dentry, mode, nd); | |
a74574aa | 1599 | if (!error) |
f38aa942 | 1600 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1601 | return error; |
1602 | } | |
1603 | ||
1604 | int may_open(struct nameidata *nd, int acc_mode, int flag) | |
1605 | { | |
4ac91378 | 1606 | struct dentry *dentry = nd->path.dentry; |
1da177e4 LT |
1607 | struct inode *inode = dentry->d_inode; |
1608 | int error; | |
1609 | ||
1610 | if (!inode) | |
1611 | return -ENOENT; | |
1612 | ||
1613 | if (S_ISLNK(inode->i_mode)) | |
1614 | return -ELOOP; | |
1615 | ||
974a9f0b | 1616 | if (S_ISDIR(inode->i_mode) && (acc_mode & MAY_WRITE)) |
1da177e4 LT |
1617 | return -EISDIR; |
1618 | ||
1da177e4 LT |
1619 | /* |
1620 | * FIFO's, sockets and device files are special: they don't | |
1621 | * actually live on the filesystem itself, and as such you | |
1622 | * can write to them even if the filesystem is read-only. | |
1623 | */ | |
1624 | if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { | |
1625 | flag &= ~O_TRUNC; | |
1626 | } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) { | |
4ac91378 | 1627 | if (nd->path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1628 | return -EACCES; |
1629 | ||
1630 | flag &= ~O_TRUNC; | |
4a3fd211 | 1631 | } |
b41572e9 DH |
1632 | |
1633 | error = vfs_permission(nd, acc_mode); | |
1634 | if (error) | |
1635 | return error; | |
1da177e4 LT |
1636 | /* |
1637 | * An append-only file must be opened in append mode for writing. | |
1638 | */ | |
1639 | if (IS_APPEND(inode)) { | |
1640 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
1641 | return -EPERM; | |
1642 | if (flag & O_TRUNC) | |
1643 | return -EPERM; | |
1644 | } | |
1645 | ||
1646 | /* O_NOATIME can only be set by the owner or superuser */ | |
1647 | if (flag & O_NOATIME) | |
3bd858ab | 1648 | if (!is_owner_or_cap(inode)) |
1da177e4 LT |
1649 | return -EPERM; |
1650 | ||
1651 | /* | |
1652 | * Ensure there are no outstanding leases on the file. | |
1653 | */ | |
1654 | error = break_lease(inode, flag); | |
1655 | if (error) | |
1656 | return error; | |
1657 | ||
1658 | if (flag & O_TRUNC) { | |
1659 | error = get_write_access(inode); | |
1660 | if (error) | |
1661 | return error; | |
1662 | ||
1663 | /* | |
1664 | * Refuse to truncate files with mandatory locks held on them. | |
1665 | */ | |
1666 | error = locks_verify_locked(inode); | |
1667 | if (!error) { | |
1668 | DQUOT_INIT(inode); | |
d139d7ff MS |
1669 | |
1670 | error = do_truncate(dentry, 0, | |
1671 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1672 | NULL); | |
1da177e4 LT |
1673 | } |
1674 | put_write_access(inode); | |
1675 | if (error) | |
1676 | return error; | |
1677 | } else | |
1678 | if (flag & FMODE_WRITE) | |
1679 | DQUOT_INIT(inode); | |
1680 | ||
1681 | return 0; | |
1682 | } | |
1683 | ||
d57999e1 DH |
1684 | /* |
1685 | * Be careful about ever adding any more callers of this | |
1686 | * function. Its flags must be in the namei format, not | |
1687 | * what get passed to sys_open(). | |
1688 | */ | |
1689 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1690 | int flag, int mode) |
1691 | { | |
1692 | int error; | |
4ac91378 | 1693 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1694 | |
1695 | if (!IS_POSIXACL(dir->d_inode)) | |
1696 | mode &= ~current->fs->umask; | |
1697 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); | |
1698 | mutex_unlock(&dir->d_inode->i_mutex); | |
4ac91378 JB |
1699 | dput(nd->path.dentry); |
1700 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1701 | if (error) |
1702 | return error; | |
1703 | /* Don't check for write permission, don't truncate */ | |
1704 | return may_open(nd, 0, flag & ~O_TRUNC); | |
1705 | } | |
1706 | ||
d57999e1 DH |
1707 | /* |
1708 | * Note that while the flag value (low two bits) for sys_open means: | |
1709 | * 00 - read-only | |
1710 | * 01 - write-only | |
1711 | * 10 - read-write | |
1712 | * 11 - special | |
1713 | * it is changed into | |
1714 | * 00 - no permissions needed | |
1715 | * 01 - read-permission | |
1716 | * 10 - write-permission | |
1717 | * 11 - read-write | |
1718 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1719 | * This is more logical, and also allows the 00 "no perm needed" | |
1720 | * to be used for symlinks (where the permissions are checked | |
1721 | * later). | |
1722 | * | |
1723 | */ | |
1724 | static inline int open_to_namei_flags(int flag) | |
1725 | { | |
1726 | if ((flag+1) & O_ACCMODE) | |
1727 | flag++; | |
1728 | return flag; | |
1729 | } | |
1730 | ||
4a3fd211 DH |
1731 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1732 | { | |
1733 | /* | |
1734 | * We'll never write to the fs underlying | |
1735 | * a device file. | |
1736 | */ | |
1737 | if (special_file(inode->i_mode)) | |
1738 | return 0; | |
1739 | return (flag & O_TRUNC); | |
1740 | } | |
1741 | ||
1da177e4 | 1742 | /* |
4a3fd211 DH |
1743 | * Note that the low bits of the passed in "open_flag" |
1744 | * are not the same as in the local variable "flag". See | |
1745 | * open_to_namei_flags() for more details. | |
1da177e4 | 1746 | */ |
a70e65df CH |
1747 | struct file *do_filp_open(int dfd, const char *pathname, |
1748 | int open_flag, int mode) | |
1da177e4 | 1749 | { |
4a3fd211 | 1750 | struct file *filp; |
a70e65df | 1751 | struct nameidata nd; |
834f2a4a | 1752 | int acc_mode, error; |
4e7506e4 | 1753 | struct path path; |
1da177e4 LT |
1754 | struct dentry *dir; |
1755 | int count = 0; | |
4a3fd211 | 1756 | int will_write; |
d57999e1 | 1757 | int flag = open_to_namei_flags(open_flag); |
1da177e4 LT |
1758 | |
1759 | acc_mode = ACC_MODE(flag); | |
1760 | ||
834f2a4a TM |
1761 | /* O_TRUNC implies we need access checks for write permissions */ |
1762 | if (flag & O_TRUNC) | |
1763 | acc_mode |= MAY_WRITE; | |
1764 | ||
1da177e4 LT |
1765 | /* Allow the LSM permission hook to distinguish append |
1766 | access from general write access. */ | |
1767 | if (flag & O_APPEND) | |
1768 | acc_mode |= MAY_APPEND; | |
1769 | ||
1da177e4 LT |
1770 | /* |
1771 | * The simplest case - just a plain lookup. | |
1772 | */ | |
1773 | if (!(flag & O_CREAT)) { | |
5590ff0d | 1774 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
a70e65df | 1775 | &nd, flag); |
1da177e4 | 1776 | if (error) |
a70e65df | 1777 | return ERR_PTR(error); |
1da177e4 LT |
1778 | goto ok; |
1779 | } | |
1780 | ||
1781 | /* | |
1782 | * Create - we need to know the parent. | |
1783 | */ | |
a70e65df CH |
1784 | error = path_lookup_create(dfd, pathname, LOOKUP_PARENT, |
1785 | &nd, flag, mode); | |
1da177e4 | 1786 | if (error) |
a70e65df | 1787 | return ERR_PTR(error); |
1da177e4 LT |
1788 | |
1789 | /* | |
1790 | * We have the parent and last component. First of all, check | |
1791 | * that we are not asked to creat(2) an obvious directory - that | |
1792 | * will not do. | |
1793 | */ | |
1794 | error = -EISDIR; | |
a70e65df | 1795 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
1da177e4 LT |
1796 | goto exit; |
1797 | ||
a70e65df CH |
1798 | dir = nd.path.dentry; |
1799 | nd.flags &= ~LOOKUP_PARENT; | |
1b1dcc1b | 1800 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1801 | path.dentry = lookup_hash(&nd); |
1802 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1803 | |
1804 | do_last: | |
4e7506e4 AV |
1805 | error = PTR_ERR(path.dentry); |
1806 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1807 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1808 | goto exit; |
1809 | } | |
1810 | ||
a70e65df | 1811 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1812 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1813 | goto exit_mutex_unlock; |
4af4c52f OD |
1814 | } |
1815 | ||
1da177e4 | 1816 | /* Negative dentry, just create the file */ |
4e7506e4 | 1817 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1818 | /* |
1819 | * This write is needed to ensure that a | |
1820 | * ro->rw transition does not occur between | |
1821 | * the time when the file is created and when | |
1822 | * a permanent write count is taken through | |
1823 | * the 'struct file' in nameidata_to_filp(). | |
1824 | */ | |
1825 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1826 | if (error) |
4a3fd211 DH |
1827 | goto exit_mutex_unlock; |
1828 | error = __open_namei_create(&nd, &path, flag, mode); | |
1829 | if (error) { | |
1830 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1831 | goto exit; |
4a3fd211 DH |
1832 | } |
1833 | filp = nameidata_to_filp(&nd, open_flag); | |
1834 | mnt_drop_write(nd.path.mnt); | |
1835 | return filp; | |
1da177e4 LT |
1836 | } |
1837 | ||
1838 | /* | |
1839 | * It already exists. | |
1840 | */ | |
1b1dcc1b | 1841 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1842 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1843 | |
1844 | error = -EEXIST; | |
1845 | if (flag & O_EXCL) | |
1846 | goto exit_dput; | |
1847 | ||
e13b210f | 1848 | if (__follow_mount(&path)) { |
1da177e4 | 1849 | error = -ELOOP; |
ba7a4c1a AV |
1850 | if (flag & O_NOFOLLOW) |
1851 | goto exit_dput; | |
1da177e4 | 1852 | } |
3e2efce0 | 1853 | |
1da177e4 | 1854 | error = -ENOENT; |
4e7506e4 | 1855 | if (!path.dentry->d_inode) |
1da177e4 | 1856 | goto exit_dput; |
4e7506e4 | 1857 | if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1858 | goto do_link; |
1859 | ||
a70e65df | 1860 | path_to_nameidata(&path, &nd); |
1da177e4 | 1861 | error = -EISDIR; |
4e7506e4 | 1862 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1863 | goto exit; |
1864 | ok: | |
4a3fd211 DH |
1865 | /* |
1866 | * Consider: | |
1867 | * 1. may_open() truncates a file | |
1868 | * 2. a rw->ro mount transition occurs | |
1869 | * 3. nameidata_to_filp() fails due to | |
1870 | * the ro mount. | |
1871 | * That would be inconsistent, and should | |
1872 | * be avoided. Taking this mnt write here | |
1873 | * ensures that (2) can not occur. | |
1874 | */ | |
1875 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1876 | if (will_write) { | |
1877 | error = mnt_want_write(nd.path.mnt); | |
1878 | if (error) | |
1879 | goto exit; | |
1880 | } | |
a70e65df | 1881 | error = may_open(&nd, acc_mode, flag); |
4a3fd211 DH |
1882 | if (error) { |
1883 | if (will_write) | |
1884 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1885 | goto exit; |
4a3fd211 DH |
1886 | } |
1887 | filp = nameidata_to_filp(&nd, open_flag); | |
1888 | /* | |
1889 | * It is now safe to drop the mnt write | |
1890 | * because the filp has had a write taken | |
1891 | * on its behalf. | |
1892 | */ | |
1893 | if (will_write) | |
1894 | mnt_drop_write(nd.path.mnt); | |
1895 | return filp; | |
1da177e4 | 1896 | |
4a3fd211 DH |
1897 | exit_mutex_unlock: |
1898 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1899 | exit_dput: |
a70e65df | 1900 | path_put_conditional(&path, &nd); |
1da177e4 | 1901 | exit: |
a70e65df CH |
1902 | if (!IS_ERR(nd.intent.open.file)) |
1903 | release_open_intent(&nd); | |
1904 | path_put(&nd.path); | |
1905 | return ERR_PTR(error); | |
1da177e4 LT |
1906 | |
1907 | do_link: | |
1908 | error = -ELOOP; | |
1909 | if (flag & O_NOFOLLOW) | |
1910 | goto exit_dput; | |
1911 | /* | |
1912 | * This is subtle. Instead of calling do_follow_link() we do the | |
1913 | * thing by hands. The reason is that this way we have zero link_count | |
1914 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1915 | * After that we have the parent and last component, i.e. | |
1916 | * we are in the same situation as after the first path_walk(). | |
1917 | * Well, almost - if the last component is normal we get its copy | |
1918 | * stored in nd->last.name and we will have to putname() it when we | |
1919 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1920 | */ | |
a70e65df CH |
1921 | nd.flags |= LOOKUP_PARENT; |
1922 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1923 | if (error) |
1924 | goto exit_dput; | |
a70e65df | 1925 | error = __do_follow_link(&path, &nd); |
de459215 KK |
1926 | if (error) { |
1927 | /* Does someone understand code flow here? Or it is only | |
1928 | * me so stupid? Anathema to whoever designed this non-sense | |
1929 | * with "intent.open". | |
1930 | */ | |
a70e65df CH |
1931 | release_open_intent(&nd); |
1932 | return ERR_PTR(error); | |
de459215 | 1933 | } |
a70e65df CH |
1934 | nd.flags &= ~LOOKUP_PARENT; |
1935 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1936 | goto ok; |
1da177e4 | 1937 | error = -EISDIR; |
a70e65df | 1938 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1939 | goto exit; |
a70e65df CH |
1940 | if (nd.last.name[nd.last.len]) { |
1941 | __putname(nd.last.name); | |
1da177e4 LT |
1942 | goto exit; |
1943 | } | |
1944 | error = -ELOOP; | |
1945 | if (count++==32) { | |
a70e65df | 1946 | __putname(nd.last.name); |
1da177e4 LT |
1947 | goto exit; |
1948 | } | |
a70e65df | 1949 | dir = nd.path.dentry; |
1b1dcc1b | 1950 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1951 | path.dentry = lookup_hash(&nd); |
1952 | path.mnt = nd.path.mnt; | |
1953 | __putname(nd.last.name); | |
1da177e4 LT |
1954 | goto do_last; |
1955 | } | |
1956 | ||
a70e65df CH |
1957 | /** |
1958 | * filp_open - open file and return file pointer | |
1959 | * | |
1960 | * @filename: path to open | |
1961 | * @flags: open flags as per the open(2) second argument | |
1962 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1963 | * | |
1964 | * This is the helper to open a file from kernelspace if you really | |
1965 | * have to. But in generally you should not do this, so please move | |
1966 | * along, nothing to see here.. | |
1967 | */ | |
1968 | struct file *filp_open(const char *filename, int flags, int mode) | |
1969 | { | |
1970 | return do_filp_open(AT_FDCWD, filename, flags, mode); | |
1971 | } | |
1972 | EXPORT_SYMBOL(filp_open); | |
1973 | ||
1da177e4 LT |
1974 | /** |
1975 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1976 | * @nd: nameidata info | |
1977 | * @is_dir: directory flag | |
1978 | * | |
1979 | * Simple function to lookup and return a dentry and create it | |
1980 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1981 | * |
4ac91378 | 1982 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1983 | */ |
1984 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1985 | { | |
c663e5d8 | 1986 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1987 | |
4ac91378 | 1988 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1989 | /* |
1990 | * Yucky last component or no last component at all? | |
1991 | * (foo/., foo/.., /////) | |
1992 | */ | |
1da177e4 LT |
1993 | if (nd->last_type != LAST_NORM) |
1994 | goto fail; | |
1995 | nd->flags &= ~LOOKUP_PARENT; | |
a634904a AM |
1996 | nd->flags |= LOOKUP_CREATE; |
1997 | nd->intent.open.flags = O_EXCL; | |
c663e5d8 CH |
1998 | |
1999 | /* | |
2000 | * Do the final lookup. | |
2001 | */ | |
49705b77 | 2002 | dentry = lookup_hash(nd); |
1da177e4 LT |
2003 | if (IS_ERR(dentry)) |
2004 | goto fail; | |
c663e5d8 CH |
2005 | |
2006 | /* | |
2007 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2008 | * From the vfs_mknod() POV we just have a negative dentry - | |
2009 | * all is fine. Let's be bastards - you had / on the end, you've | |
2010 | * been asking for (non-existent) directory. -ENOENT for you. | |
2011 | */ | |
1da177e4 LT |
2012 | if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode) |
2013 | goto enoent; | |
2014 | return dentry; | |
2015 | enoent: | |
2016 | dput(dentry); | |
2017 | dentry = ERR_PTR(-ENOENT); | |
2018 | fail: | |
2019 | return dentry; | |
2020 | } | |
f81a0bff | 2021 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
2022 | |
2023 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
2024 | { | |
2025 | int error = may_create(dir, dentry, NULL); | |
2026 | ||
2027 | if (error) | |
2028 | return error; | |
2029 | ||
2030 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
2031 | return -EPERM; | |
2032 | ||
2033 | if (!dir->i_op || !dir->i_op->mknod) | |
2034 | return -EPERM; | |
2035 | ||
08ce5f16 SH |
2036 | error = devcgroup_inode_mknod(mode, dev); |
2037 | if (error) | |
2038 | return error; | |
2039 | ||
1da177e4 LT |
2040 | error = security_inode_mknod(dir, dentry, mode, dev); |
2041 | if (error) | |
2042 | return error; | |
2043 | ||
2044 | DQUOT_INIT(dir); | |
2045 | error = dir->i_op->mknod(dir, dentry, mode, dev); | |
a74574aa | 2046 | if (!error) |
f38aa942 | 2047 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2048 | return error; |
2049 | } | |
2050 | ||
463c3197 DH |
2051 | static int may_mknod(mode_t mode) |
2052 | { | |
2053 | switch (mode & S_IFMT) { | |
2054 | case S_IFREG: | |
2055 | case S_IFCHR: | |
2056 | case S_IFBLK: | |
2057 | case S_IFIFO: | |
2058 | case S_IFSOCK: | |
2059 | case 0: /* zero mode translates to S_IFREG */ | |
2060 | return 0; | |
2061 | case S_IFDIR: | |
2062 | return -EPERM; | |
2063 | default: | |
2064 | return -EINVAL; | |
2065 | } | |
2066 | } | |
2067 | ||
5590ff0d UD |
2068 | asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode, |
2069 | unsigned dev) | |
1da177e4 LT |
2070 | { |
2071 | int error = 0; | |
2072 | char * tmp; | |
2073 | struct dentry * dentry; | |
2074 | struct nameidata nd; | |
2075 | ||
2076 | if (S_ISDIR(mode)) | |
2077 | return -EPERM; | |
2078 | tmp = getname(filename); | |
2079 | if (IS_ERR(tmp)) | |
2080 | return PTR_ERR(tmp); | |
2081 | ||
5590ff0d | 2082 | error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2083 | if (error) |
2084 | goto out; | |
2085 | dentry = lookup_create(&nd, 0); | |
463c3197 DH |
2086 | if (IS_ERR(dentry)) { |
2087 | error = PTR_ERR(dentry); | |
2088 | goto out_unlock; | |
2089 | } | |
4ac91378 | 2090 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
1da177e4 | 2091 | mode &= ~current->fs->umask; |
463c3197 DH |
2092 | error = may_mknod(mode); |
2093 | if (error) | |
2094 | goto out_dput; | |
2095 | error = mnt_want_write(nd.path.mnt); | |
2096 | if (error) | |
2097 | goto out_dput; | |
2098 | switch (mode & S_IFMT) { | |
1da177e4 | 2099 | case 0: case S_IFREG: |
4ac91378 | 2100 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2101 | break; |
2102 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2103 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2104 | new_decode_dev(dev)); |
2105 | break; | |
2106 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2107 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2108 | break; |
1da177e4 | 2109 | } |
463c3197 DH |
2110 | mnt_drop_write(nd.path.mnt); |
2111 | out_dput: | |
2112 | dput(dentry); | |
2113 | out_unlock: | |
4ac91378 | 2114 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2115 | path_put(&nd.path); |
1da177e4 LT |
2116 | out: |
2117 | putname(tmp); | |
2118 | ||
2119 | return error; | |
2120 | } | |
2121 | ||
5590ff0d UD |
2122 | asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev) |
2123 | { | |
2124 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2125 | } | |
2126 | ||
1da177e4 LT |
2127 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2128 | { | |
2129 | int error = may_create(dir, dentry, NULL); | |
2130 | ||
2131 | if (error) | |
2132 | return error; | |
2133 | ||
2134 | if (!dir->i_op || !dir->i_op->mkdir) | |
2135 | return -EPERM; | |
2136 | ||
2137 | mode &= (S_IRWXUGO|S_ISVTX); | |
2138 | error = security_inode_mkdir(dir, dentry, mode); | |
2139 | if (error) | |
2140 | return error; | |
2141 | ||
2142 | DQUOT_INIT(dir); | |
2143 | error = dir->i_op->mkdir(dir, dentry, mode); | |
a74574aa | 2144 | if (!error) |
f38aa942 | 2145 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2146 | return error; |
2147 | } | |
2148 | ||
5590ff0d | 2149 | asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode) |
1da177e4 LT |
2150 | { |
2151 | int error = 0; | |
2152 | char * tmp; | |
6902d925 DH |
2153 | struct dentry *dentry; |
2154 | struct nameidata nd; | |
1da177e4 LT |
2155 | |
2156 | tmp = getname(pathname); | |
2157 | error = PTR_ERR(tmp); | |
6902d925 DH |
2158 | if (IS_ERR(tmp)) |
2159 | goto out_err; | |
1da177e4 | 2160 | |
6902d925 DH |
2161 | error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd); |
2162 | if (error) | |
2163 | goto out; | |
2164 | dentry = lookup_create(&nd, 1); | |
2165 | error = PTR_ERR(dentry); | |
2166 | if (IS_ERR(dentry)) | |
2167 | goto out_unlock; | |
1da177e4 | 2168 | |
4ac91378 | 2169 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
6902d925 | 2170 | mode &= ~current->fs->umask; |
463c3197 DH |
2171 | error = mnt_want_write(nd.path.mnt); |
2172 | if (error) | |
2173 | goto out_dput; | |
4ac91378 | 2174 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
463c3197 DH |
2175 | mnt_drop_write(nd.path.mnt); |
2176 | out_dput: | |
6902d925 DH |
2177 | dput(dentry); |
2178 | out_unlock: | |
4ac91378 | 2179 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2180 | path_put(&nd.path); |
6902d925 DH |
2181 | out: |
2182 | putname(tmp); | |
2183 | out_err: | |
1da177e4 LT |
2184 | return error; |
2185 | } | |
2186 | ||
5590ff0d UD |
2187 | asmlinkage long sys_mkdir(const char __user *pathname, int mode) |
2188 | { | |
2189 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2190 | } | |
2191 | ||
1da177e4 LT |
2192 | /* |
2193 | * We try to drop the dentry early: we should have | |
2194 | * a usage count of 2 if we're the only user of this | |
2195 | * dentry, and if that is true (possibly after pruning | |
2196 | * the dcache), then we drop the dentry now. | |
2197 | * | |
2198 | * A low-level filesystem can, if it choses, legally | |
2199 | * do a | |
2200 | * | |
2201 | * if (!d_unhashed(dentry)) | |
2202 | * return -EBUSY; | |
2203 | * | |
2204 | * if it cannot handle the case of removing a directory | |
2205 | * that is still in use by something else.. | |
2206 | */ | |
2207 | void dentry_unhash(struct dentry *dentry) | |
2208 | { | |
2209 | dget(dentry); | |
dc168427 | 2210 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2211 | spin_lock(&dcache_lock); |
2212 | spin_lock(&dentry->d_lock); | |
2213 | if (atomic_read(&dentry->d_count) == 2) | |
2214 | __d_drop(dentry); | |
2215 | spin_unlock(&dentry->d_lock); | |
2216 | spin_unlock(&dcache_lock); | |
2217 | } | |
2218 | ||
2219 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2220 | { | |
2221 | int error = may_delete(dir, dentry, 1); | |
2222 | ||
2223 | if (error) | |
2224 | return error; | |
2225 | ||
2226 | if (!dir->i_op || !dir->i_op->rmdir) | |
2227 | return -EPERM; | |
2228 | ||
2229 | DQUOT_INIT(dir); | |
2230 | ||
1b1dcc1b | 2231 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2232 | dentry_unhash(dentry); |
2233 | if (d_mountpoint(dentry)) | |
2234 | error = -EBUSY; | |
2235 | else { | |
2236 | error = security_inode_rmdir(dir, dentry); | |
2237 | if (!error) { | |
2238 | error = dir->i_op->rmdir(dir, dentry); | |
2239 | if (!error) | |
2240 | dentry->d_inode->i_flags |= S_DEAD; | |
2241 | } | |
2242 | } | |
1b1dcc1b | 2243 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2244 | if (!error) { |
1da177e4 LT |
2245 | d_delete(dentry); |
2246 | } | |
2247 | dput(dentry); | |
2248 | ||
2249 | return error; | |
2250 | } | |
2251 | ||
5590ff0d | 2252 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2253 | { |
2254 | int error = 0; | |
2255 | char * name; | |
2256 | struct dentry *dentry; | |
2257 | struct nameidata nd; | |
2258 | ||
2259 | name = getname(pathname); | |
2260 | if(IS_ERR(name)) | |
2261 | return PTR_ERR(name); | |
2262 | ||
5590ff0d | 2263 | error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2264 | if (error) |
2265 | goto exit; | |
2266 | ||
2267 | switch(nd.last_type) { | |
2268 | case LAST_DOTDOT: | |
2269 | error = -ENOTEMPTY; | |
2270 | goto exit1; | |
2271 | case LAST_DOT: | |
2272 | error = -EINVAL; | |
2273 | goto exit1; | |
2274 | case LAST_ROOT: | |
2275 | error = -EBUSY; | |
2276 | goto exit1; | |
2277 | } | |
4ac91378 | 2278 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2279 | dentry = lookup_hash(&nd); |
1da177e4 | 2280 | error = PTR_ERR(dentry); |
6902d925 DH |
2281 | if (IS_ERR(dentry)) |
2282 | goto exit2; | |
0622753b DH |
2283 | error = mnt_want_write(nd.path.mnt); |
2284 | if (error) | |
2285 | goto exit3; | |
4ac91378 | 2286 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
0622753b DH |
2287 | mnt_drop_write(nd.path.mnt); |
2288 | exit3: | |
6902d925 DH |
2289 | dput(dentry); |
2290 | exit2: | |
4ac91378 | 2291 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2292 | exit1: |
1d957f9b | 2293 | path_put(&nd.path); |
1da177e4 LT |
2294 | exit: |
2295 | putname(name); | |
2296 | return error; | |
2297 | } | |
2298 | ||
5590ff0d UD |
2299 | asmlinkage long sys_rmdir(const char __user *pathname) |
2300 | { | |
2301 | return do_rmdir(AT_FDCWD, pathname); | |
2302 | } | |
2303 | ||
1da177e4 LT |
2304 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2305 | { | |
2306 | int error = may_delete(dir, dentry, 0); | |
2307 | ||
2308 | if (error) | |
2309 | return error; | |
2310 | ||
2311 | if (!dir->i_op || !dir->i_op->unlink) | |
2312 | return -EPERM; | |
2313 | ||
2314 | DQUOT_INIT(dir); | |
2315 | ||
1b1dcc1b | 2316 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2317 | if (d_mountpoint(dentry)) |
2318 | error = -EBUSY; | |
2319 | else { | |
2320 | error = security_inode_unlink(dir, dentry); | |
2321 | if (!error) | |
2322 | error = dir->i_op->unlink(dir, dentry); | |
2323 | } | |
1b1dcc1b | 2324 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2325 | |
2326 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2327 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2328 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2329 | d_delete(dentry); |
1da177e4 | 2330 | } |
0eeca283 | 2331 | |
1da177e4 LT |
2332 | return error; |
2333 | } | |
2334 | ||
2335 | /* | |
2336 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2337 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2338 | * writeout happening, and we don't want to prevent access to the directory |
2339 | * while waiting on the I/O. | |
2340 | */ | |
5590ff0d | 2341 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 LT |
2342 | { |
2343 | int error = 0; | |
2344 | char * name; | |
2345 | struct dentry *dentry; | |
2346 | struct nameidata nd; | |
2347 | struct inode *inode = NULL; | |
2348 | ||
2349 | name = getname(pathname); | |
2350 | if(IS_ERR(name)) | |
2351 | return PTR_ERR(name); | |
2352 | ||
5590ff0d | 2353 | error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2354 | if (error) |
2355 | goto exit; | |
2356 | error = -EISDIR; | |
2357 | if (nd.last_type != LAST_NORM) | |
2358 | goto exit1; | |
4ac91378 | 2359 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2360 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2361 | error = PTR_ERR(dentry); |
2362 | if (!IS_ERR(dentry)) { | |
2363 | /* Why not before? Because we want correct error value */ | |
2364 | if (nd.last.name[nd.last.len]) | |
2365 | goto slashes; | |
2366 | inode = dentry->d_inode; | |
2367 | if (inode) | |
2368 | atomic_inc(&inode->i_count); | |
0622753b DH |
2369 | error = mnt_want_write(nd.path.mnt); |
2370 | if (error) | |
2371 | goto exit2; | |
4ac91378 | 2372 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
0622753b | 2373 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2374 | exit2: |
2375 | dput(dentry); | |
2376 | } | |
4ac91378 | 2377 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2378 | if (inode) |
2379 | iput(inode); /* truncate the inode here */ | |
2380 | exit1: | |
1d957f9b | 2381 | path_put(&nd.path); |
1da177e4 LT |
2382 | exit: |
2383 | putname(name); | |
2384 | return error; | |
2385 | ||
2386 | slashes: | |
2387 | error = !dentry->d_inode ? -ENOENT : | |
2388 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2389 | goto exit2; | |
2390 | } | |
2391 | ||
5590ff0d UD |
2392 | asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag) |
2393 | { | |
2394 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2395 | return -EINVAL; | |
2396 | ||
2397 | if (flag & AT_REMOVEDIR) | |
2398 | return do_rmdir(dfd, pathname); | |
2399 | ||
2400 | return do_unlinkat(dfd, pathname); | |
2401 | } | |
2402 | ||
2403 | asmlinkage long sys_unlink(const char __user *pathname) | |
2404 | { | |
2405 | return do_unlinkat(AT_FDCWD, pathname); | |
2406 | } | |
2407 | ||
1da177e4 LT |
2408 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname, int mode) |
2409 | { | |
2410 | int error = may_create(dir, dentry, NULL); | |
2411 | ||
2412 | if (error) | |
2413 | return error; | |
2414 | ||
2415 | if (!dir->i_op || !dir->i_op->symlink) | |
2416 | return -EPERM; | |
2417 | ||
2418 | error = security_inode_symlink(dir, dentry, oldname); | |
2419 | if (error) | |
2420 | return error; | |
2421 | ||
2422 | DQUOT_INIT(dir); | |
2423 | error = dir->i_op->symlink(dir, dentry, oldname); | |
a74574aa | 2424 | if (!error) |
f38aa942 | 2425 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2426 | return error; |
2427 | } | |
2428 | ||
5590ff0d UD |
2429 | asmlinkage long sys_symlinkat(const char __user *oldname, |
2430 | int newdfd, const char __user *newname) | |
1da177e4 LT |
2431 | { |
2432 | int error = 0; | |
2433 | char * from; | |
2434 | char * to; | |
6902d925 DH |
2435 | struct dentry *dentry; |
2436 | struct nameidata nd; | |
1da177e4 LT |
2437 | |
2438 | from = getname(oldname); | |
2439 | if(IS_ERR(from)) | |
2440 | return PTR_ERR(from); | |
2441 | to = getname(newname); | |
2442 | error = PTR_ERR(to); | |
6902d925 DH |
2443 | if (IS_ERR(to)) |
2444 | goto out_putname; | |
1da177e4 | 2445 | |
6902d925 DH |
2446 | error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd); |
2447 | if (error) | |
2448 | goto out; | |
2449 | dentry = lookup_create(&nd, 0); | |
2450 | error = PTR_ERR(dentry); | |
2451 | if (IS_ERR(dentry)) | |
2452 | goto out_unlock; | |
2453 | ||
75c3f29d DH |
2454 | error = mnt_want_write(nd.path.mnt); |
2455 | if (error) | |
2456 | goto out_dput; | |
4ac91378 | 2457 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from, S_IALLUGO); |
75c3f29d DH |
2458 | mnt_drop_write(nd.path.mnt); |
2459 | out_dput: | |
6902d925 DH |
2460 | dput(dentry); |
2461 | out_unlock: | |
4ac91378 | 2462 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2463 | path_put(&nd.path); |
1da177e4 | 2464 | out: |
6902d925 DH |
2465 | putname(to); |
2466 | out_putname: | |
1da177e4 LT |
2467 | putname(from); |
2468 | return error; | |
2469 | } | |
2470 | ||
5590ff0d UD |
2471 | asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname) |
2472 | { | |
2473 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2474 | } | |
2475 | ||
1da177e4 LT |
2476 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2477 | { | |
2478 | struct inode *inode = old_dentry->d_inode; | |
2479 | int error; | |
2480 | ||
2481 | if (!inode) | |
2482 | return -ENOENT; | |
2483 | ||
2484 | error = may_create(dir, new_dentry, NULL); | |
2485 | if (error) | |
2486 | return error; | |
2487 | ||
2488 | if (dir->i_sb != inode->i_sb) | |
2489 | return -EXDEV; | |
2490 | ||
2491 | /* | |
2492 | * A link to an append-only or immutable file cannot be created. | |
2493 | */ | |
2494 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2495 | return -EPERM; | |
2496 | if (!dir->i_op || !dir->i_op->link) | |
2497 | return -EPERM; | |
2498 | if (S_ISDIR(old_dentry->d_inode->i_mode)) | |
2499 | return -EPERM; | |
2500 | ||
2501 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2502 | if (error) | |
2503 | return error; | |
2504 | ||
1b1dcc1b | 2505 | mutex_lock(&old_dentry->d_inode->i_mutex); |
1da177e4 LT |
2506 | DQUOT_INIT(dir); |
2507 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
1b1dcc1b | 2508 | mutex_unlock(&old_dentry->d_inode->i_mutex); |
e31e14ec | 2509 | if (!error) |
ece95912 | 2510 | fsnotify_link(dir, old_dentry->d_inode, new_dentry); |
1da177e4 LT |
2511 | return error; |
2512 | } | |
2513 | ||
2514 | /* | |
2515 | * Hardlinks are often used in delicate situations. We avoid | |
2516 | * security-related surprises by not following symlinks on the | |
2517 | * newname. --KAB | |
2518 | * | |
2519 | * We don't follow them on the oldname either to be compatible | |
2520 | * with linux 2.0, and to avoid hard-linking to directories | |
2521 | * and other special files. --ADM | |
2522 | */ | |
5590ff0d | 2523 | asmlinkage long sys_linkat(int olddfd, const char __user *oldname, |
c04030e1 UD |
2524 | int newdfd, const char __user *newname, |
2525 | int flags) | |
1da177e4 LT |
2526 | { |
2527 | struct dentry *new_dentry; | |
2528 | struct nameidata nd, old_nd; | |
2529 | int error; | |
2530 | char * to; | |
2531 | ||
45c9b11a | 2532 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2533 | return -EINVAL; |
2534 | ||
1da177e4 LT |
2535 | to = getname(newname); |
2536 | if (IS_ERR(to)) | |
2537 | return PTR_ERR(to); | |
2538 | ||
45c9b11a UD |
2539 | error = __user_walk_fd(olddfd, oldname, |
2540 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2541 | &old_nd); | |
1da177e4 LT |
2542 | if (error) |
2543 | goto exit; | |
5590ff0d | 2544 | error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2545 | if (error) |
2546 | goto out; | |
2547 | error = -EXDEV; | |
4ac91378 | 2548 | if (old_nd.path.mnt != nd.path.mnt) |
1da177e4 LT |
2549 | goto out_release; |
2550 | new_dentry = lookup_create(&nd, 0); | |
2551 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2552 | if (IS_ERR(new_dentry)) |
2553 | goto out_unlock; | |
75c3f29d DH |
2554 | error = mnt_want_write(nd.path.mnt); |
2555 | if (error) | |
2556 | goto out_dput; | |
4ac91378 | 2557 | error = vfs_link(old_nd.path.dentry, nd.path.dentry->d_inode, new_dentry); |
75c3f29d DH |
2558 | mnt_drop_write(nd.path.mnt); |
2559 | out_dput: | |
6902d925 DH |
2560 | dput(new_dentry); |
2561 | out_unlock: | |
4ac91378 | 2562 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2563 | out_release: |
1d957f9b | 2564 | path_put(&nd.path); |
1da177e4 | 2565 | out: |
1d957f9b | 2566 | path_put(&old_nd.path); |
1da177e4 LT |
2567 | exit: |
2568 | putname(to); | |
2569 | ||
2570 | return error; | |
2571 | } | |
2572 | ||
5590ff0d UD |
2573 | asmlinkage long sys_link(const char __user *oldname, const char __user *newname) |
2574 | { | |
c04030e1 | 2575 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2576 | } |
2577 | ||
1da177e4 LT |
2578 | /* |
2579 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2580 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2581 | * Problems: | |
2582 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2583 | * b) race potential - two innocent renames can create a loop together. | |
2584 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2585 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2586 | * story. |
2587 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2588 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2589 | * whether the target exists). Solution: try to be smart with locking |
2590 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2591 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2592 | * move will be locked. Thus we can rank directories by the tree |
2593 | * (ancestors first) and rank all non-directories after them. | |
2594 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2595 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2596 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2597 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2598 | * we'd better make sure that there's no link(2) for them. | |
2599 | * d) some filesystems don't support opened-but-unlinked directories, | |
2600 | * either because of layout or because they are not ready to deal with | |
2601 | * all cases correctly. The latter will be fixed (taking this sort of | |
2602 | * stuff into VFS), but the former is not going away. Solution: the same | |
2603 | * trick as in rmdir(). | |
2604 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2605 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2606 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2607 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2608 | * locking]. |
2609 | */ | |
75c96f85 AB |
2610 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2611 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2612 | { |
2613 | int error = 0; | |
2614 | struct inode *target; | |
2615 | ||
2616 | /* | |
2617 | * If we are going to change the parent - check write permissions, | |
2618 | * we'll need to flip '..'. | |
2619 | */ | |
2620 | if (new_dir != old_dir) { | |
2621 | error = permission(old_dentry->d_inode, MAY_WRITE, NULL); | |
2622 | if (error) | |
2623 | return error; | |
2624 | } | |
2625 | ||
2626 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2627 | if (error) | |
2628 | return error; | |
2629 | ||
2630 | target = new_dentry->d_inode; | |
2631 | if (target) { | |
1b1dcc1b | 2632 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2633 | dentry_unhash(new_dentry); |
2634 | } | |
2635 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2636 | error = -EBUSY; | |
2637 | else | |
2638 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2639 | if (target) { | |
2640 | if (!error) | |
2641 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2642 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2643 | if (d_unhashed(new_dentry)) |
2644 | d_rehash(new_dentry); | |
2645 | dput(new_dentry); | |
2646 | } | |
e31e14ec | 2647 | if (!error) |
349457cc MF |
2648 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2649 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2650 | return error; |
2651 | } | |
2652 | ||
75c96f85 AB |
2653 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2654 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2655 | { |
2656 | struct inode *target; | |
2657 | int error; | |
2658 | ||
2659 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2660 | if (error) | |
2661 | return error; | |
2662 | ||
2663 | dget(new_dentry); | |
2664 | target = new_dentry->d_inode; | |
2665 | if (target) | |
1b1dcc1b | 2666 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2667 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2668 | error = -EBUSY; | |
2669 | else | |
2670 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2671 | if (!error) { | |
349457cc | 2672 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2673 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2674 | } |
2675 | if (target) | |
1b1dcc1b | 2676 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2677 | dput(new_dentry); |
2678 | return error; | |
2679 | } | |
2680 | ||
2681 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2682 | struct inode *new_dir, struct dentry *new_dentry) | |
2683 | { | |
2684 | int error; | |
2685 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2686 | const char *old_name; |
1da177e4 LT |
2687 | |
2688 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2689 | return 0; | |
2690 | ||
2691 | error = may_delete(old_dir, old_dentry, is_dir); | |
2692 | if (error) | |
2693 | return error; | |
2694 | ||
2695 | if (!new_dentry->d_inode) | |
2696 | error = may_create(new_dir, new_dentry, NULL); | |
2697 | else | |
2698 | error = may_delete(new_dir, new_dentry, is_dir); | |
2699 | if (error) | |
2700 | return error; | |
2701 | ||
2702 | if (!old_dir->i_op || !old_dir->i_op->rename) | |
2703 | return -EPERM; | |
2704 | ||
2705 | DQUOT_INIT(old_dir); | |
2706 | DQUOT_INIT(new_dir); | |
2707 | ||
0eeca283 RL |
2708 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2709 | ||
1da177e4 LT |
2710 | if (is_dir) |
2711 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2712 | else | |
2713 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2714 | if (!error) { | |
0eeca283 | 2715 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2716 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2717 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2718 | } |
0eeca283 RL |
2719 | fsnotify_oldname_free(old_name); |
2720 | ||
1da177e4 LT |
2721 | return error; |
2722 | } | |
2723 | ||
5590ff0d UD |
2724 | static int do_rename(int olddfd, const char *oldname, |
2725 | int newdfd, const char *newname) | |
1da177e4 LT |
2726 | { |
2727 | int error = 0; | |
2728 | struct dentry * old_dir, * new_dir; | |
2729 | struct dentry * old_dentry, *new_dentry; | |
2730 | struct dentry * trap; | |
2731 | struct nameidata oldnd, newnd; | |
2732 | ||
5590ff0d | 2733 | error = do_path_lookup(olddfd, oldname, LOOKUP_PARENT, &oldnd); |
1da177e4 LT |
2734 | if (error) |
2735 | goto exit; | |
2736 | ||
5590ff0d | 2737 | error = do_path_lookup(newdfd, newname, LOOKUP_PARENT, &newnd); |
1da177e4 LT |
2738 | if (error) |
2739 | goto exit1; | |
2740 | ||
2741 | error = -EXDEV; | |
4ac91378 | 2742 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2743 | goto exit2; |
2744 | ||
4ac91378 | 2745 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2746 | error = -EBUSY; |
2747 | if (oldnd.last_type != LAST_NORM) | |
2748 | goto exit2; | |
2749 | ||
4ac91378 | 2750 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2751 | if (newnd.last_type != LAST_NORM) |
2752 | goto exit2; | |
2753 | ||
2754 | trap = lock_rename(new_dir, old_dir); | |
2755 | ||
49705b77 | 2756 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2757 | error = PTR_ERR(old_dentry); |
2758 | if (IS_ERR(old_dentry)) | |
2759 | goto exit3; | |
2760 | /* source must exist */ | |
2761 | error = -ENOENT; | |
2762 | if (!old_dentry->d_inode) | |
2763 | goto exit4; | |
2764 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2765 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2766 | error = -ENOTDIR; | |
2767 | if (oldnd.last.name[oldnd.last.len]) | |
2768 | goto exit4; | |
2769 | if (newnd.last.name[newnd.last.len]) | |
2770 | goto exit4; | |
2771 | } | |
2772 | /* source should not be ancestor of target */ | |
2773 | error = -EINVAL; | |
2774 | if (old_dentry == trap) | |
2775 | goto exit4; | |
49705b77 | 2776 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2777 | error = PTR_ERR(new_dentry); |
2778 | if (IS_ERR(new_dentry)) | |
2779 | goto exit4; | |
2780 | /* target should not be an ancestor of source */ | |
2781 | error = -ENOTEMPTY; | |
2782 | if (new_dentry == trap) | |
2783 | goto exit5; | |
2784 | ||
9079b1eb DH |
2785 | error = mnt_want_write(oldnd.path.mnt); |
2786 | if (error) | |
2787 | goto exit5; | |
1da177e4 LT |
2788 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2789 | new_dir->d_inode, new_dentry); | |
9079b1eb | 2790 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2791 | exit5: |
2792 | dput(new_dentry); | |
2793 | exit4: | |
2794 | dput(old_dentry); | |
2795 | exit3: | |
2796 | unlock_rename(new_dir, old_dir); | |
2797 | exit2: | |
1d957f9b | 2798 | path_put(&newnd.path); |
1da177e4 | 2799 | exit1: |
1d957f9b | 2800 | path_put(&oldnd.path); |
1da177e4 LT |
2801 | exit: |
2802 | return error; | |
2803 | } | |
2804 | ||
5590ff0d UD |
2805 | asmlinkage long sys_renameat(int olddfd, const char __user *oldname, |
2806 | int newdfd, const char __user *newname) | |
1da177e4 LT |
2807 | { |
2808 | int error; | |
2809 | char * from; | |
2810 | char * to; | |
2811 | ||
2812 | from = getname(oldname); | |
2813 | if(IS_ERR(from)) | |
2814 | return PTR_ERR(from); | |
2815 | to = getname(newname); | |
2816 | error = PTR_ERR(to); | |
2817 | if (!IS_ERR(to)) { | |
5590ff0d | 2818 | error = do_rename(olddfd, from, newdfd, to); |
1da177e4 LT |
2819 | putname(to); |
2820 | } | |
2821 | putname(from); | |
2822 | return error; | |
2823 | } | |
2824 | ||
5590ff0d UD |
2825 | asmlinkage long sys_rename(const char __user *oldname, const char __user *newname) |
2826 | { | |
2827 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2828 | } | |
2829 | ||
1da177e4 LT |
2830 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2831 | { | |
2832 | int len; | |
2833 | ||
2834 | len = PTR_ERR(link); | |
2835 | if (IS_ERR(link)) | |
2836 | goto out; | |
2837 | ||
2838 | len = strlen(link); | |
2839 | if (len > (unsigned) buflen) | |
2840 | len = buflen; | |
2841 | if (copy_to_user(buffer, link, len)) | |
2842 | len = -EFAULT; | |
2843 | out: | |
2844 | return len; | |
2845 | } | |
2846 | ||
2847 | /* | |
2848 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2849 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2850 | * using) it for any given inode is up to filesystem. | |
2851 | */ | |
2852 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2853 | { | |
2854 | struct nameidata nd; | |
cc314eef LT |
2855 | void *cookie; |
2856 | ||
1da177e4 | 2857 | nd.depth = 0; |
cc314eef LT |
2858 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
2859 | if (!IS_ERR(cookie)) { | |
2860 | int res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
1da177e4 | 2861 | if (dentry->d_inode->i_op->put_link) |
cc314eef LT |
2862 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); |
2863 | cookie = ERR_PTR(res); | |
1da177e4 | 2864 | } |
cc314eef | 2865 | return PTR_ERR(cookie); |
1da177e4 LT |
2866 | } |
2867 | ||
2868 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2869 | { | |
2870 | return __vfs_follow_link(nd, link); | |
2871 | } | |
2872 | ||
2873 | /* get the link contents into pagecache */ | |
2874 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2875 | { | |
2876 | struct page * page; | |
2877 | struct address_space *mapping = dentry->d_inode->i_mapping; | |
090d2b18 | 2878 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2879 | if (IS_ERR(page)) |
6fe6900e | 2880 | return (char*)page; |
1da177e4 LT |
2881 | *ppage = page; |
2882 | return kmap(page); | |
1da177e4 LT |
2883 | } |
2884 | ||
2885 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2886 | { | |
2887 | struct page *page = NULL; | |
2888 | char *s = page_getlink(dentry, &page); | |
2889 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2890 | if (page) { | |
2891 | kunmap(page); | |
2892 | page_cache_release(page); | |
2893 | } | |
2894 | return res; | |
2895 | } | |
2896 | ||
cc314eef | 2897 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2898 | { |
cc314eef | 2899 | struct page *page = NULL; |
1da177e4 | 2900 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2901 | return page; |
1da177e4 LT |
2902 | } |
2903 | ||
cc314eef | 2904 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2905 | { |
cc314eef LT |
2906 | struct page *page = cookie; |
2907 | ||
2908 | if (page) { | |
1da177e4 LT |
2909 | kunmap(page); |
2910 | page_cache_release(page); | |
1da177e4 LT |
2911 | } |
2912 | } | |
2913 | ||
0adb25d2 KK |
2914 | int __page_symlink(struct inode *inode, const char *symname, int len, |
2915 | gfp_t gfp_mask) | |
1da177e4 LT |
2916 | { |
2917 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2918 | struct page *page; |
afddba49 | 2919 | void *fsdata; |
beb497ab | 2920 | int err; |
1da177e4 LT |
2921 | char *kaddr; |
2922 | ||
7e53cac4 | 2923 | retry: |
afddba49 NP |
2924 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
2925 | AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata); | |
1da177e4 | 2926 | if (err) |
afddba49 NP |
2927 | goto fail; |
2928 | ||
1da177e4 LT |
2929 | kaddr = kmap_atomic(page, KM_USER0); |
2930 | memcpy(kaddr, symname, len-1); | |
2931 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2932 | |
2933 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2934 | page, fsdata); | |
1da177e4 LT |
2935 | if (err < 0) |
2936 | goto fail; | |
afddba49 NP |
2937 | if (err < len-1) |
2938 | goto retry; | |
2939 | ||
1da177e4 LT |
2940 | mark_inode_dirty(inode); |
2941 | return 0; | |
1da177e4 LT |
2942 | fail: |
2943 | return err; | |
2944 | } | |
2945 | ||
0adb25d2 KK |
2946 | int page_symlink(struct inode *inode, const char *symname, int len) |
2947 | { | |
2948 | return __page_symlink(inode, symname, len, | |
2949 | mapping_gfp_mask(inode->i_mapping)); | |
2950 | } | |
2951 | ||
92e1d5be | 2952 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2953 | .readlink = generic_readlink, |
2954 | .follow_link = page_follow_link_light, | |
2955 | .put_link = page_put_link, | |
2956 | }; | |
2957 | ||
2958 | EXPORT_SYMBOL(__user_walk); | |
5590ff0d | 2959 | EXPORT_SYMBOL(__user_walk_fd); |
1da177e4 LT |
2960 | EXPORT_SYMBOL(follow_down); |
2961 | EXPORT_SYMBOL(follow_up); | |
2962 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2963 | EXPORT_SYMBOL(getname); | |
2964 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2965 | EXPORT_SYMBOL(lookup_one_len); |
2966 | EXPORT_SYMBOL(page_follow_link_light); | |
2967 | EXPORT_SYMBOL(page_put_link); | |
2968 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2969 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2970 | EXPORT_SYMBOL(page_symlink); |
2971 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2972 | EXPORT_SYMBOL(path_lookup); | |
16f18200 | 2973 | EXPORT_SYMBOL(vfs_path_lookup); |
1da177e4 | 2974 | EXPORT_SYMBOL(permission); |
e4543edd | 2975 | EXPORT_SYMBOL(vfs_permission); |
8c744fb8 | 2976 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2977 | EXPORT_SYMBOL(unlock_rename); |
2978 | EXPORT_SYMBOL(vfs_create); | |
2979 | EXPORT_SYMBOL(vfs_follow_link); | |
2980 | EXPORT_SYMBOL(vfs_link); | |
2981 | EXPORT_SYMBOL(vfs_mkdir); | |
2982 | EXPORT_SYMBOL(vfs_mknod); | |
2983 | EXPORT_SYMBOL(generic_permission); | |
2984 | EXPORT_SYMBOL(vfs_readlink); | |
2985 | EXPORT_SYMBOL(vfs_rename); | |
2986 | EXPORT_SYMBOL(vfs_rmdir); | |
2987 | EXPORT_SYMBOL(vfs_symlink); | |
2988 | EXPORT_SYMBOL(vfs_unlink); | |
2989 | EXPORT_SYMBOL(dentry_unhash); | |
2990 | EXPORT_SYMBOL(generic_readlink); |