]>
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> |
1da177e4 LT |
33 | #include <asm/namei.h> |
34 | #include <asm/uaccess.h> | |
35 | ||
36 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
37 | ||
38 | /* [Feb-1997 T. Schoebel-Theuer] | |
39 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
40 | * were necessary because of omirr. The reason is that omirr needs | |
41 | * to know the _real_ pathname, not the user-supplied one, in case | |
42 | * of symlinks (and also when transname replacements occur). | |
43 | * | |
44 | * The new code replaces the old recursive symlink resolution with | |
45 | * an iterative one (in case of non-nested symlink chains). It does | |
46 | * this with calls to <fs>_follow_link(). | |
47 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
48 | * replaced with a single function lookup_dentry() that can handle all | |
49 | * the special cases of the former code. | |
50 | * | |
51 | * With the new dcache, the pathname is stored at each inode, at least as | |
52 | * long as the refcount of the inode is positive. As a side effect, the | |
53 | * size of the dcache depends on the inode cache and thus is dynamic. | |
54 | * | |
55 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
56 | * resolution to correspond with current state of the code. | |
57 | * | |
58 | * Note that the symlink resolution is not *completely* iterative. | |
59 | * There is still a significant amount of tail- and mid- recursion in | |
60 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
61 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
62 | * may return different results than <fs>_follow_link(). Many virtual | |
63 | * filesystems (including /proc) exhibit this behavior. | |
64 | */ | |
65 | ||
66 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
67 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
68 | * and the name already exists in form of a symlink, try to create the new | |
69 | * name indicated by the symlink. The old code always complained that the | |
70 | * name already exists, due to not following the symlink even if its target | |
71 | * is nonexistent. The new semantics affects also mknod() and link() when | |
72 | * the name is a symlink pointing to a non-existant name. | |
73 | * | |
74 | * I don't know which semantics is the right one, since I have no access | |
75 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
76 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
77 | * "old" one. Personally, I think the new semantics is much more logical. | |
78 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
79 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
80 | * and in the old Linux semantics. | |
81 | */ | |
82 | ||
83 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
84 | * semantics. See the comments in "open_namei" and "do_link" below. | |
85 | * | |
86 | * [10-Sep-98 Alan Modra] Another symlink change. | |
87 | */ | |
88 | ||
89 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
90 | * inside the path - always follow. | |
91 | * in the last component in creation/removal/renaming - never follow. | |
92 | * if LOOKUP_FOLLOW passed - follow. | |
93 | * if the pathname has trailing slashes - follow. | |
94 | * otherwise - don't follow. | |
95 | * (applied in that order). | |
96 | * | |
97 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
98 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
99 | * During the 2.4 we need to fix the userland stuff depending on it - | |
100 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
101 | * XEmacs seems to be relying on it... | |
102 | */ | |
103 | /* | |
104 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 105 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
106 | * any extra contention... |
107 | */ | |
108 | ||
fc9b52cd | 109 | static int link_path_walk(const char *name, struct nameidata *nd); |
c4a7808f | 110 | |
1da177e4 LT |
111 | /* In order to reduce some races, while at the same time doing additional |
112 | * checking and hopefully speeding things up, we copy filenames to the | |
113 | * kernel data space before using them.. | |
114 | * | |
115 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
116 | * PATH_MAX includes the nul terminator --RR. | |
117 | */ | |
858119e1 | 118 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
119 | { |
120 | int retval; | |
121 | unsigned long len = PATH_MAX; | |
122 | ||
123 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
124 | if ((unsigned long) filename >= TASK_SIZE) | |
125 | return -EFAULT; | |
126 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
127 | len = TASK_SIZE - (unsigned long) filename; | |
128 | } | |
129 | ||
130 | retval = strncpy_from_user(page, filename, len); | |
131 | if (retval > 0) { | |
132 | if (retval < len) | |
133 | return 0; | |
134 | return -ENAMETOOLONG; | |
135 | } else if (!retval) | |
136 | retval = -ENOENT; | |
137 | return retval; | |
138 | } | |
139 | ||
140 | char * getname(const char __user * filename) | |
141 | { | |
142 | char *tmp, *result; | |
143 | ||
144 | result = ERR_PTR(-ENOMEM); | |
145 | tmp = __getname(); | |
146 | if (tmp) { | |
147 | int retval = do_getname(filename, tmp); | |
148 | ||
149 | result = tmp; | |
150 | if (retval < 0) { | |
151 | __putname(tmp); | |
152 | result = ERR_PTR(retval); | |
153 | } | |
154 | } | |
155 | audit_getname(result); | |
156 | return result; | |
157 | } | |
158 | ||
159 | #ifdef CONFIG_AUDITSYSCALL | |
160 | void putname(const char *name) | |
161 | { | |
5ac3a9c2 | 162 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
163 | audit_putname(name); |
164 | else | |
165 | __putname(name); | |
166 | } | |
167 | EXPORT_SYMBOL(putname); | |
168 | #endif | |
169 | ||
170 | ||
171 | /** | |
172 | * generic_permission - check for access rights on a Posix-like filesystem | |
173 | * @inode: inode to check access rights for | |
174 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
175 | * @check_acl: optional callback to check for Posix ACLs | |
176 | * | |
177 | * Used to check for read/write/execute permissions on a file. | |
178 | * We use "fsuid" for this, letting us set arbitrary permissions | |
179 | * for filesystem access without changing the "normal" uids which | |
180 | * are used for other things.. | |
181 | */ | |
182 | int generic_permission(struct inode *inode, int mask, | |
183 | int (*check_acl)(struct inode *inode, int mask)) | |
184 | { | |
185 | umode_t mode = inode->i_mode; | |
186 | ||
187 | if (current->fsuid == inode->i_uid) | |
188 | mode >>= 6; | |
189 | else { | |
190 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
191 | int error = check_acl(inode, mask); | |
192 | if (error == -EACCES) | |
193 | goto check_capabilities; | |
194 | else if (error != -EAGAIN) | |
195 | return error; | |
196 | } | |
197 | ||
198 | if (in_group_p(inode->i_gid)) | |
199 | mode >>= 3; | |
200 | } | |
201 | ||
202 | /* | |
203 | * If the DACs are ok we don't need any capability check. | |
204 | */ | |
205 | if (((mode & mask & (MAY_READ|MAY_WRITE|MAY_EXEC)) == mask)) | |
206 | return 0; | |
207 | ||
208 | check_capabilities: | |
209 | /* | |
210 | * Read/write DACs are always overridable. | |
211 | * Executable DACs are overridable if at least one exec bit is set. | |
212 | */ | |
213 | if (!(mask & MAY_EXEC) || | |
214 | (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode)) | |
215 | if (capable(CAP_DAC_OVERRIDE)) | |
216 | return 0; | |
217 | ||
218 | /* | |
219 | * Searching includes executable on directories, else just read. | |
220 | */ | |
221 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
222 | if (capable(CAP_DAC_READ_SEARCH)) | |
223 | return 0; | |
224 | ||
225 | return -EACCES; | |
226 | } | |
227 | ||
228 | int permission(struct inode *inode, int mask, struct nameidata *nd) | |
229 | { | |
230 | int retval, submask; | |
c7eb2667 DH |
231 | struct vfsmount *mnt = NULL; |
232 | ||
233 | if (nd) | |
4ac91378 | 234 | mnt = nd->path.mnt; |
1da177e4 LT |
235 | |
236 | if (mask & MAY_WRITE) { | |
22590e41 | 237 | umode_t mode = inode->i_mode; |
1da177e4 LT |
238 | |
239 | /* | |
240 | * Nobody gets write access to a read-only fs. | |
241 | */ | |
242 | if (IS_RDONLY(inode) && | |
243 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
244 | return -EROFS; | |
245 | ||
246 | /* | |
247 | * Nobody gets write access to an immutable file. | |
248 | */ | |
249 | if (IS_IMMUTABLE(inode)) | |
250 | return -EACCES; | |
251 | } | |
252 | ||
22590e41 MS |
253 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) { |
254 | /* | |
255 | * MAY_EXEC on regular files is denied if the fs is mounted | |
256 | * with the "noexec" flag. | |
257 | */ | |
c7eb2667 | 258 | if (mnt && (mnt->mnt_flags & MNT_NOEXEC)) |
22590e41 MS |
259 | return -EACCES; |
260 | } | |
a343bb77 | 261 | |
1da177e4 LT |
262 | /* Ordinary permission routines do not understand MAY_APPEND. */ |
263 | submask = mask & ~MAY_APPEND; | |
22590e41 | 264 | if (inode->i_op && inode->i_op->permission) { |
1da177e4 | 265 | retval = inode->i_op->permission(inode, submask, nd); |
22590e41 MS |
266 | if (!retval) { |
267 | /* | |
268 | * Exec permission on a regular file is denied if none | |
269 | * of the execute bits are set. | |
270 | * | |
271 | * This check should be done by the ->permission() | |
272 | * method. | |
273 | */ | |
274 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode) && | |
275 | !(inode->i_mode & S_IXUGO)) | |
276 | return -EACCES; | |
277 | } | |
278 | } else { | |
1da177e4 | 279 | retval = generic_permission(inode, submask, NULL); |
22590e41 | 280 | } |
1da177e4 LT |
281 | if (retval) |
282 | return retval; | |
283 | ||
284 | return security_inode_permission(inode, mask, nd); | |
285 | } | |
286 | ||
e4543edd CH |
287 | /** |
288 | * vfs_permission - check for access rights to a given path | |
289 | * @nd: lookup result that describes the path | |
290 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
291 | * | |
292 | * Used to check for read/write/execute permissions on a path. | |
293 | * We use "fsuid" for this, letting us set arbitrary permissions | |
294 | * for filesystem access without changing the "normal" uids which | |
295 | * are used for other things. | |
296 | */ | |
297 | int vfs_permission(struct nameidata *nd, int mask) | |
298 | { | |
4ac91378 | 299 | return permission(nd->path.dentry->d_inode, mask, nd); |
e4543edd CH |
300 | } |
301 | ||
8c744fb8 CH |
302 | /** |
303 | * file_permission - check for additional access rights to a given file | |
304 | * @file: file to check access rights for | |
305 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
306 | * | |
307 | * Used to check for read/write/execute permissions on an already opened | |
308 | * file. | |
309 | * | |
310 | * Note: | |
311 | * Do not use this function in new code. All access checks should | |
312 | * be done using vfs_permission(). | |
313 | */ | |
314 | int file_permission(struct file *file, int mask) | |
315 | { | |
0f7fc9e4 | 316 | return permission(file->f_path.dentry->d_inode, mask, NULL); |
8c744fb8 CH |
317 | } |
318 | ||
1da177e4 LT |
319 | /* |
320 | * get_write_access() gets write permission for a file. | |
321 | * put_write_access() releases this write permission. | |
322 | * This is used for regular files. | |
323 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
324 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
325 | * can have the following values: | |
326 | * 0: no writers, no VM_DENYWRITE mappings | |
327 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
328 | * > 0: (i_writecount) users are writing to the file. | |
329 | * | |
330 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
331 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
332 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
333 | * to do the change if sign is wrong. Exclusion between them is provided by | |
334 | * the inode->i_lock spinlock. | |
335 | */ | |
336 | ||
337 | int get_write_access(struct inode * inode) | |
338 | { | |
339 | spin_lock(&inode->i_lock); | |
340 | if (atomic_read(&inode->i_writecount) < 0) { | |
341 | spin_unlock(&inode->i_lock); | |
342 | return -ETXTBSY; | |
343 | } | |
344 | atomic_inc(&inode->i_writecount); | |
345 | spin_unlock(&inode->i_lock); | |
346 | ||
347 | return 0; | |
348 | } | |
349 | ||
350 | int deny_write_access(struct file * file) | |
351 | { | |
0f7fc9e4 | 352 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
353 | |
354 | spin_lock(&inode->i_lock); | |
355 | if (atomic_read(&inode->i_writecount) > 0) { | |
356 | spin_unlock(&inode->i_lock); | |
357 | return -ETXTBSY; | |
358 | } | |
359 | atomic_dec(&inode->i_writecount); | |
360 | spin_unlock(&inode->i_lock); | |
361 | ||
362 | return 0; | |
363 | } | |
364 | ||
365 | void path_release(struct nameidata *nd) | |
366 | { | |
4ac91378 JB |
367 | dput(nd->path.dentry); |
368 | mntput(nd->path.mnt); | |
1da177e4 LT |
369 | } |
370 | ||
834f2a4a TM |
371 | /** |
372 | * release_open_intent - free up open intent resources | |
373 | * @nd: pointer to nameidata | |
374 | */ | |
375 | void release_open_intent(struct nameidata *nd) | |
376 | { | |
0f7fc9e4 | 377 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
378 | put_filp(nd->intent.open.file); |
379 | else | |
380 | fput(nd->intent.open.file); | |
381 | } | |
382 | ||
bcdc5e01 IK |
383 | static inline struct dentry * |
384 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
385 | { | |
386 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
387 | if (unlikely(status <= 0)) { | |
388 | /* | |
389 | * The dentry failed validation. | |
390 | * If d_revalidate returned 0 attempt to invalidate | |
391 | * the dentry otherwise d_revalidate is asking us | |
392 | * to return a fail status. | |
393 | */ | |
394 | if (!status) { | |
395 | if (!d_invalidate(dentry)) { | |
396 | dput(dentry); | |
397 | dentry = NULL; | |
398 | } | |
399 | } else { | |
400 | dput(dentry); | |
401 | dentry = ERR_PTR(status); | |
402 | } | |
403 | } | |
404 | return dentry; | |
405 | } | |
406 | ||
1da177e4 LT |
407 | /* |
408 | * Internal lookup() using the new generic dcache. | |
409 | * SMP-safe | |
410 | */ | |
411 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
412 | { | |
413 | struct dentry * dentry = __d_lookup(parent, name); | |
414 | ||
415 | /* lockess __d_lookup may fail due to concurrent d_move() | |
416 | * in some unrelated directory, so try with d_lookup | |
417 | */ | |
418 | if (!dentry) | |
419 | dentry = d_lookup(parent, name); | |
420 | ||
bcdc5e01 IK |
421 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
422 | dentry = do_revalidate(dentry, nd); | |
423 | ||
1da177e4 LT |
424 | return dentry; |
425 | } | |
426 | ||
427 | /* | |
428 | * Short-cut version of permission(), for calling by | |
429 | * path_walk(), when dcache lock is held. Combines parts | |
430 | * of permission() and generic_permission(), and tests ONLY for | |
431 | * MAY_EXEC permission. | |
432 | * | |
433 | * If appropriate, check DAC only. If not appropriate, or | |
434 | * short-cut DAC fails, then call permission() to do more | |
435 | * complete permission check. | |
436 | */ | |
858119e1 | 437 | static int exec_permission_lite(struct inode *inode, |
1da177e4 LT |
438 | struct nameidata *nd) |
439 | { | |
440 | umode_t mode = inode->i_mode; | |
441 | ||
442 | if (inode->i_op && inode->i_op->permission) | |
443 | return -EAGAIN; | |
444 | ||
445 | if (current->fsuid == inode->i_uid) | |
446 | mode >>= 6; | |
447 | else if (in_group_p(inode->i_gid)) | |
448 | mode >>= 3; | |
449 | ||
450 | if (mode & MAY_EXEC) | |
451 | goto ok; | |
452 | ||
453 | if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE)) | |
454 | goto ok; | |
455 | ||
456 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE)) | |
457 | goto ok; | |
458 | ||
459 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH)) | |
460 | goto ok; | |
461 | ||
462 | return -EACCES; | |
463 | ok: | |
464 | return security_inode_permission(inode, MAY_EXEC, nd); | |
465 | } | |
466 | ||
467 | /* | |
468 | * This is called when everything else fails, and we actually have | |
469 | * to go to the low-level filesystem to find out what we should do.. | |
470 | * | |
471 | * We get the directory semaphore, and after getting that we also | |
472 | * make sure that nobody added the entry to the dcache in the meantime.. | |
473 | * SMP-safe | |
474 | */ | |
475 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
476 | { | |
477 | struct dentry * result; | |
478 | struct inode *dir = parent->d_inode; | |
479 | ||
1b1dcc1b | 480 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
481 | /* |
482 | * First re-do the cached lookup just in case it was created | |
483 | * while we waited for the directory semaphore.. | |
484 | * | |
485 | * FIXME! This could use version numbering or similar to | |
486 | * avoid unnecessary cache lookups. | |
487 | * | |
488 | * The "dcache_lock" is purely to protect the RCU list walker | |
489 | * from concurrent renames at this point (we mustn't get false | |
490 | * negatives from the RCU list walk here, unlike the optimistic | |
491 | * fast walk). | |
492 | * | |
493 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
494 | */ | |
495 | result = d_lookup(parent, name); | |
496 | if (!result) { | |
497 | struct dentry * dentry = d_alloc(parent, name); | |
498 | result = ERR_PTR(-ENOMEM); | |
499 | if (dentry) { | |
500 | result = dir->i_op->lookup(dir, dentry, nd); | |
501 | if (result) | |
502 | dput(dentry); | |
503 | else | |
504 | result = dentry; | |
505 | } | |
1b1dcc1b | 506 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
507 | return result; |
508 | } | |
509 | ||
510 | /* | |
511 | * Uhhuh! Nasty case: the cache was re-populated while | |
512 | * we waited on the semaphore. Need to revalidate. | |
513 | */ | |
1b1dcc1b | 514 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 515 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
516 | result = do_revalidate(result, nd); |
517 | if (!result) | |
1da177e4 | 518 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
519 | } |
520 | return result; | |
521 | } | |
522 | ||
523 | static int __emul_lookup_dentry(const char *, struct nameidata *); | |
524 | ||
525 | /* SMP-safe */ | |
f1662356 | 526 | static __always_inline int |
1da177e4 LT |
527 | walk_init_root(const char *name, struct nameidata *nd) |
528 | { | |
e518ddb7 AM |
529 | struct fs_struct *fs = current->fs; |
530 | ||
531 | read_lock(&fs->lock); | |
532 | if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) { | |
4ac91378 JB |
533 | nd->path.mnt = mntget(fs->altrootmnt); |
534 | nd->path.dentry = dget(fs->altroot); | |
e518ddb7 | 535 | read_unlock(&fs->lock); |
1da177e4 LT |
536 | if (__emul_lookup_dentry(name,nd)) |
537 | return 0; | |
e518ddb7 | 538 | read_lock(&fs->lock); |
1da177e4 | 539 | } |
4ac91378 JB |
540 | nd->path.mnt = mntget(fs->rootmnt); |
541 | nd->path.dentry = dget(fs->root); | |
e518ddb7 | 542 | read_unlock(&fs->lock); |
1da177e4 LT |
543 | return 1; |
544 | } | |
545 | ||
f1662356 | 546 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
547 | { |
548 | int res = 0; | |
549 | char *name; | |
550 | if (IS_ERR(link)) | |
551 | goto fail; | |
552 | ||
553 | if (*link == '/') { | |
554 | path_release(nd); | |
555 | if (!walk_init_root(link, nd)) | |
556 | /* weird __emul_prefix() stuff did it */ | |
557 | goto out; | |
558 | } | |
559 | res = link_path_walk(link, nd); | |
560 | out: | |
561 | if (nd->depth || res || nd->last_type!=LAST_NORM) | |
562 | return res; | |
563 | /* | |
564 | * If it is an iterative symlinks resolution in open_namei() we | |
565 | * have to copy the last component. And all that crap because of | |
566 | * bloody create() on broken symlinks. Furrfu... | |
567 | */ | |
568 | name = __getname(); | |
569 | if (unlikely(!name)) { | |
570 | path_release(nd); | |
571 | return -ENOMEM; | |
572 | } | |
573 | strcpy(name, nd->last.name); | |
574 | nd->last.name = name; | |
575 | return 0; | |
576 | fail: | |
577 | path_release(nd); | |
578 | return PTR_ERR(link); | |
579 | } | |
580 | ||
051d3812 IK |
581 | static inline void dput_path(struct path *path, struct nameidata *nd) |
582 | { | |
583 | dput(path->dentry); | |
4ac91378 | 584 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
585 | mntput(path->mnt); |
586 | } | |
587 | ||
588 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
589 | { | |
4ac91378 JB |
590 | dput(nd->path.dentry); |
591 | if (nd->path.mnt != path->mnt) | |
592 | mntput(nd->path.mnt); | |
593 | nd->path.mnt = path->mnt; | |
594 | nd->path.dentry = path->dentry; | |
051d3812 IK |
595 | } |
596 | ||
f1662356 | 597 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
598 | { |
599 | int error; | |
cc314eef | 600 | void *cookie; |
cd4e91d3 | 601 | struct dentry *dentry = path->dentry; |
1da177e4 | 602 | |
d671a1cb | 603 | touch_atime(path->mnt, dentry); |
1da177e4 | 604 | nd_set_link(nd, NULL); |
cd4e91d3 | 605 | |
4ac91378 | 606 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
607 | path_to_nameidata(path, nd); |
608 | dget(dentry); | |
609 | } | |
610 | mntget(path->mnt); | |
cc314eef LT |
611 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
612 | error = PTR_ERR(cookie); | |
613 | if (!IS_ERR(cookie)) { | |
1da177e4 | 614 | char *s = nd_get_link(nd); |
cc314eef | 615 | error = 0; |
1da177e4 LT |
616 | if (s) |
617 | error = __vfs_follow_link(nd, s); | |
618 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 619 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 620 | } |
cd4e91d3 AV |
621 | dput(dentry); |
622 | mntput(path->mnt); | |
1da177e4 LT |
623 | |
624 | return error; | |
625 | } | |
626 | ||
627 | /* | |
628 | * This limits recursive symlink follows to 8, while | |
629 | * limiting consecutive symlinks to 40. | |
630 | * | |
631 | * Without that kind of total limit, nasty chains of consecutive | |
632 | * symlinks can cause almost arbitrarily long lookups. | |
633 | */ | |
90ebe565 | 634 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
635 | { |
636 | int err = -ELOOP; | |
637 | if (current->link_count >= MAX_NESTED_LINKS) | |
638 | goto loop; | |
639 | if (current->total_link_count >= 40) | |
640 | goto loop; | |
641 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
642 | cond_resched(); | |
90ebe565 | 643 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
644 | if (err) |
645 | goto loop; | |
646 | current->link_count++; | |
647 | current->total_link_count++; | |
648 | nd->depth++; | |
cd4e91d3 | 649 | err = __do_follow_link(path, nd); |
839d9f93 AV |
650 | current->link_count--; |
651 | nd->depth--; | |
1da177e4 LT |
652 | return err; |
653 | loop: | |
09dd17d3 | 654 | dput_path(path, nd); |
839d9f93 | 655 | path_release(nd); |
1da177e4 LT |
656 | return err; |
657 | } | |
658 | ||
659 | int follow_up(struct vfsmount **mnt, struct dentry **dentry) | |
660 | { | |
661 | struct vfsmount *parent; | |
662 | struct dentry *mountpoint; | |
663 | spin_lock(&vfsmount_lock); | |
664 | parent=(*mnt)->mnt_parent; | |
665 | if (parent == *mnt) { | |
666 | spin_unlock(&vfsmount_lock); | |
667 | return 0; | |
668 | } | |
669 | mntget(parent); | |
670 | mountpoint=dget((*mnt)->mnt_mountpoint); | |
671 | spin_unlock(&vfsmount_lock); | |
672 | dput(*dentry); | |
673 | *dentry = mountpoint; | |
674 | mntput(*mnt); | |
675 | *mnt = parent; | |
676 | return 1; | |
677 | } | |
678 | ||
679 | /* no need for dcache_lock, as serialization is taken care in | |
680 | * namespace.c | |
681 | */ | |
463ffb2e AV |
682 | static int __follow_mount(struct path *path) |
683 | { | |
684 | int res = 0; | |
685 | while (d_mountpoint(path->dentry)) { | |
686 | struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry); | |
687 | if (!mounted) | |
688 | break; | |
689 | dput(path->dentry); | |
690 | if (res) | |
691 | mntput(path->mnt); | |
692 | path->mnt = mounted; | |
693 | path->dentry = dget(mounted->mnt_root); | |
694 | res = 1; | |
695 | } | |
696 | return res; | |
697 | } | |
698 | ||
58c465eb | 699 | static void follow_mount(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 | 700 | { |
1da177e4 LT |
701 | while (d_mountpoint(*dentry)) { |
702 | struct vfsmount *mounted = lookup_mnt(*mnt, *dentry); | |
703 | if (!mounted) | |
704 | break; | |
58c465eb | 705 | dput(*dentry); |
1da177e4 LT |
706 | mntput(*mnt); |
707 | *mnt = mounted; | |
1da177e4 | 708 | *dentry = dget(mounted->mnt_root); |
1da177e4 | 709 | } |
1da177e4 LT |
710 | } |
711 | ||
712 | /* no need for dcache_lock, as serialization is taken care in | |
713 | * namespace.c | |
714 | */ | |
e13b210f | 715 | int follow_down(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 LT |
716 | { |
717 | struct vfsmount *mounted; | |
718 | ||
719 | mounted = lookup_mnt(*mnt, *dentry); | |
720 | if (mounted) { | |
e13b210f | 721 | dput(*dentry); |
1da177e4 LT |
722 | mntput(*mnt); |
723 | *mnt = mounted; | |
1da177e4 LT |
724 | *dentry = dget(mounted->mnt_root); |
725 | return 1; | |
726 | } | |
727 | return 0; | |
728 | } | |
729 | ||
f1662356 | 730 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 731 | { |
e518ddb7 AM |
732 | struct fs_struct *fs = current->fs; |
733 | ||
1da177e4 LT |
734 | while(1) { |
735 | struct vfsmount *parent; | |
4ac91378 | 736 | struct dentry *old = nd->path.dentry; |
1da177e4 | 737 | |
e518ddb7 | 738 | read_lock(&fs->lock); |
4ac91378 JB |
739 | if (nd->path.dentry == fs->root && |
740 | nd->path.mnt == fs->rootmnt) { | |
e518ddb7 | 741 | read_unlock(&fs->lock); |
1da177e4 LT |
742 | break; |
743 | } | |
e518ddb7 | 744 | read_unlock(&fs->lock); |
1da177e4 | 745 | spin_lock(&dcache_lock); |
4ac91378 JB |
746 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
747 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
748 | spin_unlock(&dcache_lock); |
749 | dput(old); | |
750 | break; | |
751 | } | |
752 | spin_unlock(&dcache_lock); | |
753 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
754 | parent = nd->path.mnt->mnt_parent; |
755 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
756 | spin_unlock(&vfsmount_lock); |
757 | break; | |
758 | } | |
759 | mntget(parent); | |
4ac91378 | 760 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
761 | spin_unlock(&vfsmount_lock); |
762 | dput(old); | |
4ac91378 JB |
763 | mntput(nd->path.mnt); |
764 | nd->path.mnt = parent; | |
1da177e4 | 765 | } |
4ac91378 | 766 | follow_mount(&nd->path.mnt, &nd->path.dentry); |
1da177e4 LT |
767 | } |
768 | ||
1da177e4 LT |
769 | /* |
770 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
771 | * small and for now I'd prefer to have fast path as straight as possible. | |
772 | * It _is_ time-critical. | |
773 | */ | |
774 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
775 | struct path *path) | |
776 | { | |
4ac91378 JB |
777 | struct vfsmount *mnt = nd->path.mnt; |
778 | struct dentry *dentry = __d_lookup(nd->path.dentry, name); | |
1da177e4 LT |
779 | |
780 | if (!dentry) | |
781 | goto need_lookup; | |
782 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
783 | goto need_revalidate; | |
784 | done: | |
785 | path->mnt = mnt; | |
786 | path->dentry = dentry; | |
634ee701 | 787 | __follow_mount(path); |
1da177e4 LT |
788 | return 0; |
789 | ||
790 | need_lookup: | |
4ac91378 | 791 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
792 | if (IS_ERR(dentry)) |
793 | goto fail; | |
794 | goto done; | |
795 | ||
796 | need_revalidate: | |
bcdc5e01 IK |
797 | dentry = do_revalidate(dentry, nd); |
798 | if (!dentry) | |
799 | goto need_lookup; | |
800 | if (IS_ERR(dentry)) | |
801 | goto fail; | |
802 | goto done; | |
1da177e4 LT |
803 | |
804 | fail: | |
805 | return PTR_ERR(dentry); | |
806 | } | |
807 | ||
808 | /* | |
809 | * Name resolution. | |
ea3834d9 PM |
810 | * This is the basic name resolution function, turning a pathname into |
811 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 812 | * |
ea3834d9 PM |
813 | * Returns 0 and nd will have valid dentry and mnt on success. |
814 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 815 | */ |
fc9b52cd | 816 | static int __link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
817 | { |
818 | struct path next; | |
819 | struct inode *inode; | |
820 | int err; | |
821 | unsigned int lookup_flags = nd->flags; | |
822 | ||
823 | while (*name=='/') | |
824 | name++; | |
825 | if (!*name) | |
826 | goto return_reval; | |
827 | ||
4ac91378 | 828 | inode = nd->path.dentry->d_inode; |
1da177e4 | 829 | if (nd->depth) |
f55eab82 | 830 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
831 | |
832 | /* At this point we know we have a real path component. */ | |
833 | for(;;) { | |
834 | unsigned long hash; | |
835 | struct qstr this; | |
836 | unsigned int c; | |
837 | ||
cdce5d6b | 838 | nd->flags |= LOOKUP_CONTINUE; |
1da177e4 | 839 | err = exec_permission_lite(inode, nd); |
e4543edd CH |
840 | if (err == -EAGAIN) |
841 | err = vfs_permission(nd, MAY_EXEC); | |
1da177e4 LT |
842 | if (err) |
843 | break; | |
844 | ||
845 | this.name = name; | |
846 | c = *(const unsigned char *)name; | |
847 | ||
848 | hash = init_name_hash(); | |
849 | do { | |
850 | name++; | |
851 | hash = partial_name_hash(c, hash); | |
852 | c = *(const unsigned char *)name; | |
853 | } while (c && (c != '/')); | |
854 | this.len = name - (const char *) this.name; | |
855 | this.hash = end_name_hash(hash); | |
856 | ||
857 | /* remove trailing slashes? */ | |
858 | if (!c) | |
859 | goto last_component; | |
860 | while (*++name == '/'); | |
861 | if (!*name) | |
862 | goto last_with_slashes; | |
863 | ||
864 | /* | |
865 | * "." and ".." are special - ".." especially so because it has | |
866 | * to be able to know about the current root directory and | |
867 | * parent relationships. | |
868 | */ | |
869 | if (this.name[0] == '.') switch (this.len) { | |
870 | default: | |
871 | break; | |
872 | case 2: | |
873 | if (this.name[1] != '.') | |
874 | break; | |
58c465eb | 875 | follow_dotdot(nd); |
4ac91378 | 876 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
877 | /* fallthrough */ |
878 | case 1: | |
879 | continue; | |
880 | } | |
881 | /* | |
882 | * See if the low-level filesystem might want | |
883 | * to use its own hash.. | |
884 | */ | |
4ac91378 JB |
885 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
886 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
887 | &this); | |
1da177e4 LT |
888 | if (err < 0) |
889 | break; | |
890 | } | |
1da177e4 LT |
891 | /* This does the actual lookups.. */ |
892 | err = do_lookup(nd, &this, &next); | |
893 | if (err) | |
894 | break; | |
1da177e4 LT |
895 | |
896 | err = -ENOENT; | |
897 | inode = next.dentry->d_inode; | |
898 | if (!inode) | |
899 | goto out_dput; | |
900 | err = -ENOTDIR; | |
901 | if (!inode->i_op) | |
902 | goto out_dput; | |
903 | ||
904 | if (inode->i_op->follow_link) { | |
90ebe565 | 905 | err = do_follow_link(&next, nd); |
1da177e4 LT |
906 | if (err) |
907 | goto return_err; | |
908 | err = -ENOENT; | |
4ac91378 | 909 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
910 | if (!inode) |
911 | break; | |
912 | err = -ENOTDIR; | |
913 | if (!inode->i_op) | |
914 | break; | |
09dd17d3 MS |
915 | } else |
916 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
917 | err = -ENOTDIR; |
918 | if (!inode->i_op->lookup) | |
919 | break; | |
920 | continue; | |
921 | /* here ends the main loop */ | |
922 | ||
923 | last_with_slashes: | |
924 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
925 | last_component: | |
f55eab82 TM |
926 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
927 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
928 | if (lookup_flags & LOOKUP_PARENT) |
929 | goto lookup_parent; | |
930 | if (this.name[0] == '.') switch (this.len) { | |
931 | default: | |
932 | break; | |
933 | case 2: | |
934 | if (this.name[1] != '.') | |
935 | break; | |
58c465eb | 936 | follow_dotdot(nd); |
4ac91378 | 937 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
938 | /* fallthrough */ |
939 | case 1: | |
940 | goto return_reval; | |
941 | } | |
4ac91378 JB |
942 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
943 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
944 | &this); | |
1da177e4 LT |
945 | if (err < 0) |
946 | break; | |
947 | } | |
948 | err = do_lookup(nd, &this, &next); | |
949 | if (err) | |
950 | break; | |
1da177e4 LT |
951 | inode = next.dentry->d_inode; |
952 | if ((lookup_flags & LOOKUP_FOLLOW) | |
953 | && inode && inode->i_op && inode->i_op->follow_link) { | |
90ebe565 | 954 | err = do_follow_link(&next, nd); |
1da177e4 LT |
955 | if (err) |
956 | goto return_err; | |
4ac91378 | 957 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
958 | } else |
959 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
960 | err = -ENOENT; |
961 | if (!inode) | |
962 | break; | |
963 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
964 | err = -ENOTDIR; | |
965 | if (!inode->i_op || !inode->i_op->lookup) | |
966 | break; | |
967 | } | |
968 | goto return_base; | |
969 | lookup_parent: | |
970 | nd->last = this; | |
971 | nd->last_type = LAST_NORM; | |
972 | if (this.name[0] != '.') | |
973 | goto return_base; | |
974 | if (this.len == 1) | |
975 | nd->last_type = LAST_DOT; | |
976 | else if (this.len == 2 && this.name[1] == '.') | |
977 | nd->last_type = LAST_DOTDOT; | |
978 | else | |
979 | goto return_base; | |
980 | return_reval: | |
981 | /* | |
982 | * We bypassed the ordinary revalidation routines. | |
983 | * We may need to check the cached dentry for staleness. | |
984 | */ | |
4ac91378 JB |
985 | if (nd->path.dentry && nd->path.dentry->d_sb && |
986 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
987 | err = -ESTALE; |
988 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
989 | if (!nd->path.dentry->d_op->d_revalidate( |
990 | nd->path.dentry, nd)) | |
1da177e4 LT |
991 | break; |
992 | } | |
993 | return_base: | |
994 | return 0; | |
995 | out_dput: | |
09dd17d3 | 996 | dput_path(&next, nd); |
1da177e4 LT |
997 | break; |
998 | } | |
999 | path_release(nd); | |
1000 | return_err: | |
1001 | return err; | |
1002 | } | |
1003 | ||
1004 | /* | |
1005 | * Wrapper to retry pathname resolution whenever the underlying | |
1006 | * file system returns an ESTALE. | |
1007 | * | |
1008 | * Retry the whole path once, forcing real lookup requests | |
1009 | * instead of relying on the dcache. | |
1010 | */ | |
fc9b52cd | 1011 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1012 | { |
1013 | struct nameidata save = *nd; | |
1014 | int result; | |
1015 | ||
1016 | /* make sure the stuff we saved doesn't go away */ | |
4ac91378 JB |
1017 | dget(save.path.dentry); |
1018 | mntget(save.path.mnt); | |
1da177e4 LT |
1019 | |
1020 | result = __link_path_walk(name, nd); | |
1021 | if (result == -ESTALE) { | |
1022 | *nd = save; | |
4ac91378 JB |
1023 | dget(nd->path.dentry); |
1024 | mntget(nd->path.mnt); | |
1da177e4 LT |
1025 | nd->flags |= LOOKUP_REVAL; |
1026 | result = __link_path_walk(name, nd); | |
1027 | } | |
1028 | ||
4ac91378 JB |
1029 | dput(save.path.dentry); |
1030 | mntput(save.path.mnt); | |
1da177e4 LT |
1031 | |
1032 | return result; | |
1033 | } | |
1034 | ||
fc9b52cd | 1035 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1036 | { |
1037 | current->total_link_count = 0; | |
1038 | return link_path_walk(name, nd); | |
1039 | } | |
1040 | ||
ea3834d9 PM |
1041 | /* |
1042 | * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if | |
1043 | * everything is done. Returns 0 and drops input nd, if lookup failed; | |
1044 | */ | |
1da177e4 LT |
1045 | static int __emul_lookup_dentry(const char *name, struct nameidata *nd) |
1046 | { | |
1047 | if (path_walk(name, nd)) | |
1048 | return 0; /* something went wrong... */ | |
1049 | ||
4ac91378 JB |
1050 | if (!nd->path.dentry->d_inode || |
1051 | S_ISDIR(nd->path.dentry->d_inode->i_mode)) { | |
1052 | struct dentry *old_dentry = nd->path.dentry; | |
1053 | struct vfsmount *old_mnt = nd->path.mnt; | |
1da177e4 LT |
1054 | struct qstr last = nd->last; |
1055 | int last_type = nd->last_type; | |
e518ddb7 AM |
1056 | struct fs_struct *fs = current->fs; |
1057 | ||
1da177e4 | 1058 | /* |
e518ddb7 AM |
1059 | * NAME was not found in alternate root or it's a directory. |
1060 | * Try to find it in the normal root: | |
1da177e4 LT |
1061 | */ |
1062 | nd->last_type = LAST_ROOT; | |
e518ddb7 | 1063 | read_lock(&fs->lock); |
4ac91378 JB |
1064 | nd->path.mnt = mntget(fs->rootmnt); |
1065 | nd->path.dentry = dget(fs->root); | |
e518ddb7 | 1066 | read_unlock(&fs->lock); |
1da177e4 | 1067 | if (path_walk(name, nd) == 0) { |
4ac91378 | 1068 | if (nd->path.dentry->d_inode) { |
1da177e4 LT |
1069 | dput(old_dentry); |
1070 | mntput(old_mnt); | |
1071 | return 1; | |
1072 | } | |
1073 | path_release(nd); | |
1074 | } | |
4ac91378 JB |
1075 | nd->path.dentry = old_dentry; |
1076 | nd->path.mnt = old_mnt; | |
1da177e4 LT |
1077 | nd->last = last; |
1078 | nd->last_type = last_type; | |
1079 | } | |
1080 | return 1; | |
1081 | } | |
1082 | ||
1083 | void set_fs_altroot(void) | |
1084 | { | |
1085 | char *emul = __emul_prefix(); | |
1086 | struct nameidata nd; | |
1087 | struct vfsmount *mnt = NULL, *oldmnt; | |
1088 | struct dentry *dentry = NULL, *olddentry; | |
1089 | int err; | |
e518ddb7 | 1090 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1091 | |
1092 | if (!emul) | |
1093 | goto set_it; | |
1094 | err = path_lookup(emul, LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_NOALT, &nd); | |
1095 | if (!err) { | |
4ac91378 JB |
1096 | mnt = nd.path.mnt; |
1097 | dentry = nd.path.dentry; | |
1da177e4 LT |
1098 | } |
1099 | set_it: | |
e518ddb7 AM |
1100 | write_lock(&fs->lock); |
1101 | oldmnt = fs->altrootmnt; | |
1102 | olddentry = fs->altroot; | |
1103 | fs->altrootmnt = mnt; | |
1104 | fs->altroot = dentry; | |
1105 | write_unlock(&fs->lock); | |
1da177e4 LT |
1106 | if (olddentry) { |
1107 | dput(olddentry); | |
1108 | mntput(oldmnt); | |
1109 | } | |
1110 | } | |
1111 | ||
ea3834d9 | 1112 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
fc9b52cd | 1113 | static int do_path_lookup(int dfd, const char *name, |
5590ff0d | 1114 | unsigned int flags, struct nameidata *nd) |
1da177e4 | 1115 | { |
ea3834d9 | 1116 | int retval = 0; |
170aa3d0 UD |
1117 | int fput_needed; |
1118 | struct file *file; | |
e518ddb7 | 1119 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1120 | |
1121 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1122 | nd->flags = flags; | |
1123 | nd->depth = 0; | |
1124 | ||
1da177e4 | 1125 | if (*name=='/') { |
e518ddb7 AM |
1126 | read_lock(&fs->lock); |
1127 | if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) { | |
4ac91378 JB |
1128 | nd->path.mnt = mntget(fs->altrootmnt); |
1129 | nd->path.dentry = dget(fs->altroot); | |
e518ddb7 | 1130 | read_unlock(&fs->lock); |
1da177e4 | 1131 | if (__emul_lookup_dentry(name,nd)) |
ea3834d9 | 1132 | goto out; /* found in altroot */ |
e518ddb7 | 1133 | read_lock(&fs->lock); |
1da177e4 | 1134 | } |
4ac91378 JB |
1135 | nd->path.mnt = mntget(fs->rootmnt); |
1136 | nd->path.dentry = dget(fs->root); | |
e518ddb7 | 1137 | read_unlock(&fs->lock); |
5590ff0d | 1138 | } else if (dfd == AT_FDCWD) { |
e518ddb7 | 1139 | read_lock(&fs->lock); |
4ac91378 JB |
1140 | nd->path.mnt = mntget(fs->pwdmnt); |
1141 | nd->path.dentry = dget(fs->pwd); | |
e518ddb7 | 1142 | read_unlock(&fs->lock); |
5590ff0d | 1143 | } else { |
5590ff0d UD |
1144 | struct dentry *dentry; |
1145 | ||
1146 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1147 | retval = -EBADF; |
1148 | if (!file) | |
6d09bb62 | 1149 | goto out_fail; |
5590ff0d | 1150 | |
0f7fc9e4 | 1151 | dentry = file->f_path.dentry; |
5590ff0d | 1152 | |
170aa3d0 UD |
1153 | retval = -ENOTDIR; |
1154 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1155 | goto fput_fail; |
5590ff0d UD |
1156 | |
1157 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1158 | if (retval) |
6d09bb62 | 1159 | goto fput_fail; |
5590ff0d | 1160 | |
4ac91378 JB |
1161 | nd->path.mnt = mntget(file->f_path.mnt); |
1162 | nd->path.dentry = dget(dentry); | |
5590ff0d UD |
1163 | |
1164 | fput_light(file, fput_needed); | |
1da177e4 | 1165 | } |
2dfdd266 JJS |
1166 | |
1167 | retval = path_walk(name, nd); | |
ea3834d9 | 1168 | out: |
4ac91378 JB |
1169 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1170 | nd->path.dentry->d_inode)) | |
1171 | audit_inode(name, nd->path.dentry); | |
6d09bb62 | 1172 | out_fail: |
170aa3d0 UD |
1173 | return retval; |
1174 | ||
6d09bb62 | 1175 | fput_fail: |
170aa3d0 | 1176 | fput_light(file, fput_needed); |
6d09bb62 | 1177 | goto out_fail; |
1da177e4 LT |
1178 | } |
1179 | ||
fc9b52cd | 1180 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1181 | struct nameidata *nd) |
1182 | { | |
1183 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1184 | } | |
1185 | ||
16f18200 JJS |
1186 | /** |
1187 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1188 | * @dentry: pointer to dentry of the base directory | |
1189 | * @mnt: pointer to vfs mount of the base directory | |
1190 | * @name: pointer to file name | |
1191 | * @flags: lookup flags | |
1192 | * @nd: pointer to nameidata | |
1193 | */ | |
1194 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1195 | const char *name, unsigned int flags, | |
1196 | struct nameidata *nd) | |
1197 | { | |
1198 | int retval; | |
1199 | ||
1200 | /* same as do_path_lookup */ | |
1201 | nd->last_type = LAST_ROOT; | |
1202 | nd->flags = flags; | |
1203 | nd->depth = 0; | |
1204 | ||
4ac91378 JB |
1205 | nd->path.mnt = mntget(mnt); |
1206 | nd->path.dentry = dget(dentry); | |
16f18200 JJS |
1207 | |
1208 | retval = path_walk(name, nd); | |
4ac91378 JB |
1209 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1210 | nd->path.dentry->d_inode)) | |
1211 | audit_inode(name, nd->path.dentry); | |
16f18200 JJS |
1212 | |
1213 | return retval; | |
1214 | ||
1215 | } | |
1216 | ||
5590ff0d UD |
1217 | static int __path_lookup_intent_open(int dfd, const char *name, |
1218 | unsigned int lookup_flags, struct nameidata *nd, | |
1219 | int open_flags, int create_mode) | |
834f2a4a TM |
1220 | { |
1221 | struct file *filp = get_empty_filp(); | |
1222 | int err; | |
1223 | ||
1224 | if (filp == NULL) | |
1225 | return -ENFILE; | |
1226 | nd->intent.open.file = filp; | |
1227 | nd->intent.open.flags = open_flags; | |
1228 | nd->intent.open.create_mode = create_mode; | |
5590ff0d | 1229 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1230 | if (IS_ERR(nd->intent.open.file)) { |
1231 | if (err == 0) { | |
1232 | err = PTR_ERR(nd->intent.open.file); | |
1233 | path_release(nd); | |
1234 | } | |
1235 | } else if (err != 0) | |
1236 | release_open_intent(nd); | |
1237 | return err; | |
1238 | } | |
1239 | ||
1240 | /** | |
1241 | * path_lookup_open - lookup a file path with open intent | |
7045f37b | 1242 | * @dfd: the directory to use as base, or AT_FDCWD |
834f2a4a TM |
1243 | * @name: pointer to file name |
1244 | * @lookup_flags: lookup intent flags | |
1245 | * @nd: pointer to nameidata | |
1246 | * @open_flags: open intent flags | |
1247 | */ | |
5590ff0d | 1248 | int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags, |
834f2a4a TM |
1249 | struct nameidata *nd, int open_flags) |
1250 | { | |
5590ff0d | 1251 | return __path_lookup_intent_open(dfd, name, lookup_flags, nd, |
834f2a4a TM |
1252 | open_flags, 0); |
1253 | } | |
1254 | ||
1255 | /** | |
1256 | * path_lookup_create - lookup a file path with open + create intent | |
7045f37b | 1257 | * @dfd: the directory to use as base, or AT_FDCWD |
834f2a4a TM |
1258 | * @name: pointer to file name |
1259 | * @lookup_flags: lookup intent flags | |
1260 | * @nd: pointer to nameidata | |
1261 | * @open_flags: open intent flags | |
1262 | * @create_mode: create intent flags | |
1263 | */ | |
5590ff0d UD |
1264 | static int path_lookup_create(int dfd, const char *name, |
1265 | unsigned int lookup_flags, struct nameidata *nd, | |
1266 | int open_flags, int create_mode) | |
834f2a4a | 1267 | { |
5590ff0d UD |
1268 | return __path_lookup_intent_open(dfd, name, lookup_flags|LOOKUP_CREATE, |
1269 | nd, open_flags, create_mode); | |
834f2a4a TM |
1270 | } |
1271 | ||
1272 | int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags, | |
1273 | struct nameidata *nd, int open_flags) | |
1274 | { | |
1275 | char *tmp = getname(name); | |
1276 | int err = PTR_ERR(tmp); | |
1277 | ||
1278 | if (!IS_ERR(tmp)) { | |
5590ff0d | 1279 | err = __path_lookup_intent_open(AT_FDCWD, tmp, lookup_flags, nd, open_flags, 0); |
834f2a4a TM |
1280 | putname(tmp); |
1281 | } | |
1282 | return err; | |
1283 | } | |
1284 | ||
eead1911 CH |
1285 | static struct dentry *__lookup_hash(struct qstr *name, |
1286 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1287 | { |
057f6c01 | 1288 | struct dentry *dentry; |
1da177e4 LT |
1289 | struct inode *inode; |
1290 | int err; | |
1291 | ||
1292 | inode = base->d_inode; | |
1da177e4 LT |
1293 | |
1294 | /* | |
1295 | * See if the low-level filesystem might want | |
1296 | * to use its own hash.. | |
1297 | */ | |
1298 | if (base->d_op && base->d_op->d_hash) { | |
1299 | err = base->d_op->d_hash(base, name); | |
1300 | dentry = ERR_PTR(err); | |
1301 | if (err < 0) | |
1302 | goto out; | |
1303 | } | |
1304 | ||
1305 | dentry = cached_lookup(base, name, nd); | |
1306 | if (!dentry) { | |
1307 | struct dentry *new = d_alloc(base, name); | |
1308 | dentry = ERR_PTR(-ENOMEM); | |
1309 | if (!new) | |
1310 | goto out; | |
1311 | dentry = inode->i_op->lookup(inode, new, nd); | |
1312 | if (!dentry) | |
1313 | dentry = new; | |
1314 | else | |
1315 | dput(new); | |
1316 | } | |
1317 | out: | |
1318 | return dentry; | |
1319 | } | |
1320 | ||
057f6c01 JM |
1321 | /* |
1322 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1323 | * needs parent already locked. Doesn't follow mounts. | |
1324 | * SMP-safe. | |
1325 | */ | |
eead1911 | 1326 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1327 | { |
057f6c01 JM |
1328 | int err; |
1329 | ||
4ac91378 | 1330 | err = permission(nd->path.dentry->d_inode, MAY_EXEC, nd); |
057f6c01 | 1331 | if (err) |
eead1911 | 1332 | return ERR_PTR(err); |
4ac91378 | 1333 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1334 | } |
1335 | ||
eead1911 CH |
1336 | static int __lookup_one_len(const char *name, struct qstr *this, |
1337 | struct dentry *base, int len) | |
1da177e4 LT |
1338 | { |
1339 | unsigned long hash; | |
1da177e4 LT |
1340 | unsigned int c; |
1341 | ||
057f6c01 JM |
1342 | this->name = name; |
1343 | this->len = len; | |
1da177e4 | 1344 | if (!len) |
057f6c01 | 1345 | return -EACCES; |
1da177e4 LT |
1346 | |
1347 | hash = init_name_hash(); | |
1348 | while (len--) { | |
1349 | c = *(const unsigned char *)name++; | |
1350 | if (c == '/' || c == '\0') | |
057f6c01 | 1351 | return -EACCES; |
1da177e4 LT |
1352 | hash = partial_name_hash(c, hash); |
1353 | } | |
057f6c01 JM |
1354 | this->hash = end_name_hash(hash); |
1355 | return 0; | |
1356 | } | |
1da177e4 | 1357 | |
eead1911 CH |
1358 | /** |
1359 | * lookup_one_len: filesystem helper to lookup single pathname component | |
1360 | * @name: pathname component to lookup | |
1361 | * @base: base directory to lookup from | |
1362 | * @len: maximum length @len should be interpreted to | |
1363 | * | |
1364 | * Note that this routine is purely a helper for filesystem useage and should | |
1365 | * not be called by generic code. Also note that by using this function to | |
1366 | * nameidata argument is passed to the filesystem methods and a filesystem | |
1367 | * using this helper needs to be prepared for that. | |
1368 | */ | |
057f6c01 JM |
1369 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1370 | { | |
1371 | int err; | |
1372 | struct qstr this; | |
1373 | ||
1374 | err = __lookup_one_len(name, &this, base, len); | |
eead1911 CH |
1375 | if (err) |
1376 | return ERR_PTR(err); | |
1377 | ||
1378 | err = permission(base->d_inode, MAY_EXEC, NULL); | |
057f6c01 JM |
1379 | if (err) |
1380 | return ERR_PTR(err); | |
49705b77 | 1381 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1382 | } |
1383 | ||
eead1911 CH |
1384 | /** |
1385 | * lookup_one_noperm - bad hack for sysfs | |
1386 | * @name: pathname component to lookup | |
1387 | * @base: base directory to lookup from | |
1388 | * | |
1389 | * This is a variant of lookup_one_len that doesn't perform any permission | |
1390 | * checks. It's a horrible hack to work around the braindead sysfs | |
1391 | * architecture and should not be used anywhere else. | |
1392 | * | |
1393 | * DON'T USE THIS FUNCTION EVER, thanks. | |
1394 | */ | |
1395 | struct dentry *lookup_one_noperm(const char *name, struct dentry *base) | |
057f6c01 JM |
1396 | { |
1397 | int err; | |
1398 | struct qstr this; | |
1399 | ||
eead1911 | 1400 | err = __lookup_one_len(name, &this, base, strlen(name)); |
057f6c01 JM |
1401 | if (err) |
1402 | return ERR_PTR(err); | |
eead1911 | 1403 | return __lookup_hash(&this, base, NULL); |
1da177e4 LT |
1404 | } |
1405 | ||
fc9b52cd | 1406 | int __user_walk_fd(int dfd, const char __user *name, unsigned flags, |
5590ff0d | 1407 | struct nameidata *nd) |
1da177e4 LT |
1408 | { |
1409 | char *tmp = getname(name); | |
1410 | int err = PTR_ERR(tmp); | |
1411 | ||
1412 | if (!IS_ERR(tmp)) { | |
5590ff0d | 1413 | err = do_path_lookup(dfd, tmp, flags, nd); |
1da177e4 LT |
1414 | putname(tmp); |
1415 | } | |
1416 | return err; | |
1417 | } | |
1418 | ||
fc9b52cd | 1419 | int __user_walk(const char __user *name, unsigned flags, struct nameidata *nd) |
5590ff0d UD |
1420 | { |
1421 | return __user_walk_fd(AT_FDCWD, name, flags, nd); | |
1422 | } | |
1423 | ||
1da177e4 LT |
1424 | /* |
1425 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1426 | * minimal. | |
1427 | */ | |
1428 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1429 | { | |
1430 | if (!(dir->i_mode & S_ISVTX)) | |
1431 | return 0; | |
1432 | if (inode->i_uid == current->fsuid) | |
1433 | return 0; | |
1434 | if (dir->i_uid == current->fsuid) | |
1435 | return 0; | |
1436 | return !capable(CAP_FOWNER); | |
1437 | } | |
1438 | ||
1439 | /* | |
1440 | * Check whether we can remove a link victim from directory dir, check | |
1441 | * whether the type of victim is right. | |
1442 | * 1. We can't do it if dir is read-only (done in permission()) | |
1443 | * 2. We should have write and exec permissions on dir | |
1444 | * 3. We can't remove anything from append-only dir | |
1445 | * 4. We can't do anything with immutable dir (done in permission()) | |
1446 | * 5. If the sticky bit on dir is set we should either | |
1447 | * a. be owner of dir, or | |
1448 | * b. be owner of victim, or | |
1449 | * c. have CAP_FOWNER capability | |
1450 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1451 | * links pointing to it. | |
1452 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1453 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1454 | * 9. We can't remove a root or mountpoint. | |
1455 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1456 | * nfs_async_unlink(). | |
1457 | */ | |
858119e1 | 1458 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1459 | { |
1460 | int error; | |
1461 | ||
1462 | if (!victim->d_inode) | |
1463 | return -ENOENT; | |
1464 | ||
1465 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1466 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 LT |
1467 | |
1468 | error = permission(dir,MAY_WRITE | MAY_EXEC, NULL); | |
1469 | if (error) | |
1470 | return error; | |
1471 | if (IS_APPEND(dir)) | |
1472 | return -EPERM; | |
1473 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
1474 | IS_IMMUTABLE(victim->d_inode)) | |
1475 | return -EPERM; | |
1476 | if (isdir) { | |
1477 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1478 | return -ENOTDIR; | |
1479 | if (IS_ROOT(victim)) | |
1480 | return -EBUSY; | |
1481 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1482 | return -EISDIR; | |
1483 | if (IS_DEADDIR(dir)) | |
1484 | return -ENOENT; | |
1485 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1486 | return -EBUSY; | |
1487 | return 0; | |
1488 | } | |
1489 | ||
1490 | /* Check whether we can create an object with dentry child in directory | |
1491 | * dir. | |
1492 | * 1. We can't do it if child already exists (open has special treatment for | |
1493 | * this case, but since we are inlined it's OK) | |
1494 | * 2. We can't do it if dir is read-only (done in permission()) | |
1495 | * 3. We should have write and exec permissions on dir | |
1496 | * 4. We can't do it if dir is immutable (done in permission()) | |
1497 | */ | |
1498 | static inline int may_create(struct inode *dir, struct dentry *child, | |
1499 | struct nameidata *nd) | |
1500 | { | |
1501 | if (child->d_inode) | |
1502 | return -EEXIST; | |
1503 | if (IS_DEADDIR(dir)) | |
1504 | return -ENOENT; | |
1505 | return permission(dir,MAY_WRITE | MAY_EXEC, nd); | |
1506 | } | |
1507 | ||
1508 | /* | |
1da177e4 LT |
1509 | * O_DIRECTORY translates into forcing a directory lookup. |
1510 | */ | |
1511 | static inline int lookup_flags(unsigned int f) | |
1512 | { | |
1513 | unsigned long retval = LOOKUP_FOLLOW; | |
1514 | ||
1515 | if (f & O_NOFOLLOW) | |
1516 | retval &= ~LOOKUP_FOLLOW; | |
1517 | ||
1da177e4 LT |
1518 | if (f & O_DIRECTORY) |
1519 | retval |= LOOKUP_DIRECTORY; | |
1520 | ||
1521 | return retval; | |
1522 | } | |
1523 | ||
1524 | /* | |
1525 | * p1 and p2 should be directories on the same fs. | |
1526 | */ | |
1527 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1528 | { | |
1529 | struct dentry *p; | |
1530 | ||
1531 | if (p1 == p2) { | |
f2eace23 | 1532 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1533 | return NULL; |
1534 | } | |
1535 | ||
a11f3a05 | 1536 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1537 | |
1538 | for (p = p1; p->d_parent != p; p = p->d_parent) { | |
1539 | if (p->d_parent == p2) { | |
f2eace23 IM |
1540 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); |
1541 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1542 | return p; |
1543 | } | |
1544 | } | |
1545 | ||
1546 | for (p = p2; p->d_parent != p; p = p->d_parent) { | |
1547 | if (p->d_parent == p1) { | |
f2eace23 IM |
1548 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1549 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1550 | return p; |
1551 | } | |
1552 | } | |
1553 | ||
f2eace23 IM |
1554 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1555 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1556 | return NULL; |
1557 | } | |
1558 | ||
1559 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1560 | { | |
1b1dcc1b | 1561 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1562 | if (p1 != p2) { |
1b1dcc1b | 1563 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1564 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1565 | } |
1566 | } | |
1567 | ||
1568 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1569 | struct nameidata *nd) | |
1570 | { | |
1571 | int error = may_create(dir, dentry, nd); | |
1572 | ||
1573 | if (error) | |
1574 | return error; | |
1575 | ||
1576 | if (!dir->i_op || !dir->i_op->create) | |
1577 | return -EACCES; /* shouldn't it be ENOSYS? */ | |
1578 | mode &= S_IALLUGO; | |
1579 | mode |= S_IFREG; | |
1580 | error = security_inode_create(dir, dentry, mode); | |
1581 | if (error) | |
1582 | return error; | |
1583 | DQUOT_INIT(dir); | |
1584 | error = dir->i_op->create(dir, dentry, mode, nd); | |
a74574aa | 1585 | if (!error) |
f38aa942 | 1586 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1587 | return error; |
1588 | } | |
1589 | ||
1590 | int may_open(struct nameidata *nd, int acc_mode, int flag) | |
1591 | { | |
4ac91378 | 1592 | struct dentry *dentry = nd->path.dentry; |
1da177e4 LT |
1593 | struct inode *inode = dentry->d_inode; |
1594 | int error; | |
1595 | ||
1596 | if (!inode) | |
1597 | return -ENOENT; | |
1598 | ||
1599 | if (S_ISLNK(inode->i_mode)) | |
1600 | return -ELOOP; | |
1601 | ||
974a9f0b | 1602 | if (S_ISDIR(inode->i_mode) && (acc_mode & MAY_WRITE)) |
1da177e4 LT |
1603 | return -EISDIR; |
1604 | ||
1da177e4 LT |
1605 | /* |
1606 | * FIFO's, sockets and device files are special: they don't | |
1607 | * actually live on the filesystem itself, and as such you | |
1608 | * can write to them even if the filesystem is read-only. | |
1609 | */ | |
1610 | if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { | |
1611 | flag &= ~O_TRUNC; | |
1612 | } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) { | |
4ac91378 | 1613 | if (nd->path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1614 | return -EACCES; |
1615 | ||
1616 | flag &= ~O_TRUNC; | |
974a9f0b | 1617 | } else if (IS_RDONLY(inode) && (acc_mode & MAY_WRITE)) |
1da177e4 | 1618 | return -EROFS; |
b41572e9 DH |
1619 | |
1620 | error = vfs_permission(nd, acc_mode); | |
1621 | if (error) | |
1622 | return error; | |
1da177e4 LT |
1623 | /* |
1624 | * An append-only file must be opened in append mode for writing. | |
1625 | */ | |
1626 | if (IS_APPEND(inode)) { | |
1627 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
1628 | return -EPERM; | |
1629 | if (flag & O_TRUNC) | |
1630 | return -EPERM; | |
1631 | } | |
1632 | ||
1633 | /* O_NOATIME can only be set by the owner or superuser */ | |
1634 | if (flag & O_NOATIME) | |
3bd858ab | 1635 | if (!is_owner_or_cap(inode)) |
1da177e4 LT |
1636 | return -EPERM; |
1637 | ||
1638 | /* | |
1639 | * Ensure there are no outstanding leases on the file. | |
1640 | */ | |
1641 | error = break_lease(inode, flag); | |
1642 | if (error) | |
1643 | return error; | |
1644 | ||
1645 | if (flag & O_TRUNC) { | |
1646 | error = get_write_access(inode); | |
1647 | if (error) | |
1648 | return error; | |
1649 | ||
1650 | /* | |
1651 | * Refuse to truncate files with mandatory locks held on them. | |
1652 | */ | |
1653 | error = locks_verify_locked(inode); | |
1654 | if (!error) { | |
1655 | DQUOT_INIT(inode); | |
d139d7ff MS |
1656 | |
1657 | error = do_truncate(dentry, 0, | |
1658 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1659 | NULL); | |
1da177e4 LT |
1660 | } |
1661 | put_write_access(inode); | |
1662 | if (error) | |
1663 | return error; | |
1664 | } else | |
1665 | if (flag & FMODE_WRITE) | |
1666 | DQUOT_INIT(inode); | |
1667 | ||
1668 | return 0; | |
1669 | } | |
1670 | ||
aab520e2 DH |
1671 | static int open_namei_create(struct nameidata *nd, struct path *path, |
1672 | int flag, int mode) | |
1673 | { | |
1674 | int error; | |
4ac91378 | 1675 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1676 | |
1677 | if (!IS_POSIXACL(dir->d_inode)) | |
1678 | mode &= ~current->fs->umask; | |
1679 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); | |
1680 | mutex_unlock(&dir->d_inode->i_mutex); | |
4ac91378 JB |
1681 | dput(nd->path.dentry); |
1682 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1683 | if (error) |
1684 | return error; | |
1685 | /* Don't check for write permission, don't truncate */ | |
1686 | return may_open(nd, 0, flag & ~O_TRUNC); | |
1687 | } | |
1688 | ||
1da177e4 LT |
1689 | /* |
1690 | * open_namei() | |
1691 | * | |
1692 | * namei for open - this is in fact almost the whole open-routine. | |
1693 | * | |
1694 | * Note that the low bits of "flag" aren't the same as in the open | |
1695 | * system call - they are 00 - no permissions needed | |
1696 | * 01 - read permission needed | |
1697 | * 10 - write permission needed | |
1698 | * 11 - read/write permissions needed | |
1699 | * which is a lot more logical, and also allows the "no perm" needed | |
1700 | * for symlinks (where the permissions are checked later). | |
1701 | * SMP-safe | |
1702 | */ | |
5590ff0d UD |
1703 | int open_namei(int dfd, const char *pathname, int flag, |
1704 | int mode, struct nameidata *nd) | |
1da177e4 | 1705 | { |
834f2a4a | 1706 | int acc_mode, error; |
4e7506e4 | 1707 | struct path path; |
1da177e4 LT |
1708 | struct dentry *dir; |
1709 | int count = 0; | |
1710 | ||
1711 | acc_mode = ACC_MODE(flag); | |
1712 | ||
834f2a4a TM |
1713 | /* O_TRUNC implies we need access checks for write permissions */ |
1714 | if (flag & O_TRUNC) | |
1715 | acc_mode |= MAY_WRITE; | |
1716 | ||
1da177e4 LT |
1717 | /* Allow the LSM permission hook to distinguish append |
1718 | access from general write access. */ | |
1719 | if (flag & O_APPEND) | |
1720 | acc_mode |= MAY_APPEND; | |
1721 | ||
1da177e4 LT |
1722 | /* |
1723 | * The simplest case - just a plain lookup. | |
1724 | */ | |
1725 | if (!(flag & O_CREAT)) { | |
5590ff0d UD |
1726 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
1727 | nd, flag); | |
1da177e4 LT |
1728 | if (error) |
1729 | return error; | |
1730 | goto ok; | |
1731 | } | |
1732 | ||
1733 | /* | |
1734 | * Create - we need to know the parent. | |
1735 | */ | |
5590ff0d | 1736 | error = path_lookup_create(dfd,pathname,LOOKUP_PARENT,nd,flag,mode); |
1da177e4 LT |
1737 | if (error) |
1738 | return error; | |
1739 | ||
1740 | /* | |
1741 | * We have the parent and last component. First of all, check | |
1742 | * that we are not asked to creat(2) an obvious directory - that | |
1743 | * will not do. | |
1744 | */ | |
1745 | error = -EISDIR; | |
1746 | if (nd->last_type != LAST_NORM || nd->last.name[nd->last.len]) | |
1747 | goto exit; | |
1748 | ||
4ac91378 | 1749 | dir = nd->path.dentry; |
1da177e4 | 1750 | nd->flags &= ~LOOKUP_PARENT; |
1b1dcc1b | 1751 | mutex_lock(&dir->d_inode->i_mutex); |
49705b77 | 1752 | path.dentry = lookup_hash(nd); |
4ac91378 | 1753 | path.mnt = nd->path.mnt; |
1da177e4 LT |
1754 | |
1755 | do_last: | |
4e7506e4 AV |
1756 | error = PTR_ERR(path.dentry); |
1757 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1758 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1759 | goto exit; |
1760 | } | |
1761 | ||
4af4c52f OD |
1762 | if (IS_ERR(nd->intent.open.file)) { |
1763 | mutex_unlock(&dir->d_inode->i_mutex); | |
1764 | error = PTR_ERR(nd->intent.open.file); | |
1765 | goto exit_dput; | |
1766 | } | |
1767 | ||
1da177e4 | 1768 | /* Negative dentry, just create the file */ |
4e7506e4 | 1769 | if (!path.dentry->d_inode) { |
aab520e2 | 1770 | error = open_namei_create(nd, &path, flag, mode); |
1da177e4 LT |
1771 | if (error) |
1772 | goto exit; | |
aab520e2 | 1773 | return 0; |
1da177e4 LT |
1774 | } |
1775 | ||
1776 | /* | |
1777 | * It already exists. | |
1778 | */ | |
1b1dcc1b | 1779 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1780 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1781 | |
1782 | error = -EEXIST; | |
1783 | if (flag & O_EXCL) | |
1784 | goto exit_dput; | |
1785 | ||
e13b210f | 1786 | if (__follow_mount(&path)) { |
1da177e4 | 1787 | error = -ELOOP; |
ba7a4c1a AV |
1788 | if (flag & O_NOFOLLOW) |
1789 | goto exit_dput; | |
1da177e4 | 1790 | } |
3e2efce0 | 1791 | |
1da177e4 | 1792 | error = -ENOENT; |
4e7506e4 | 1793 | if (!path.dentry->d_inode) |
1da177e4 | 1794 | goto exit_dput; |
4e7506e4 | 1795 | if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1796 | goto do_link; |
1797 | ||
09dd17d3 | 1798 | path_to_nameidata(&path, nd); |
1da177e4 | 1799 | error = -EISDIR; |
4e7506e4 | 1800 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1801 | goto exit; |
1802 | ok: | |
1803 | error = may_open(nd, acc_mode, flag); | |
1804 | if (error) | |
1805 | goto exit; | |
1806 | return 0; | |
1807 | ||
1808 | exit_dput: | |
09dd17d3 | 1809 | dput_path(&path, nd); |
1da177e4 | 1810 | exit: |
834f2a4a TM |
1811 | if (!IS_ERR(nd->intent.open.file)) |
1812 | release_open_intent(nd); | |
1da177e4 LT |
1813 | path_release(nd); |
1814 | return error; | |
1815 | ||
1816 | do_link: | |
1817 | error = -ELOOP; | |
1818 | if (flag & O_NOFOLLOW) | |
1819 | goto exit_dput; | |
1820 | /* | |
1821 | * This is subtle. Instead of calling do_follow_link() we do the | |
1822 | * thing by hands. The reason is that this way we have zero link_count | |
1823 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1824 | * After that we have the parent and last component, i.e. | |
1825 | * we are in the same situation as after the first path_walk(). | |
1826 | * Well, almost - if the last component is normal we get its copy | |
1827 | * stored in nd->last.name and we will have to putname() it when we | |
1828 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1829 | */ | |
1830 | nd->flags |= LOOKUP_PARENT; | |
4e7506e4 | 1831 | error = security_inode_follow_link(path.dentry, nd); |
1da177e4 LT |
1832 | if (error) |
1833 | goto exit_dput; | |
cd4e91d3 | 1834 | error = __do_follow_link(&path, nd); |
de459215 KK |
1835 | if (error) { |
1836 | /* Does someone understand code flow here? Or it is only | |
1837 | * me so stupid? Anathema to whoever designed this non-sense | |
1838 | * with "intent.open". | |
1839 | */ | |
1840 | release_open_intent(nd); | |
1da177e4 | 1841 | return error; |
de459215 | 1842 | } |
1da177e4 | 1843 | nd->flags &= ~LOOKUP_PARENT; |
d671d5e5 | 1844 | if (nd->last_type == LAST_BIND) |
1da177e4 | 1845 | goto ok; |
1da177e4 LT |
1846 | error = -EISDIR; |
1847 | if (nd->last_type != LAST_NORM) | |
1848 | goto exit; | |
1849 | if (nd->last.name[nd->last.len]) { | |
82984114 | 1850 | __putname(nd->last.name); |
1da177e4 LT |
1851 | goto exit; |
1852 | } | |
1853 | error = -ELOOP; | |
1854 | if (count++==32) { | |
82984114 | 1855 | __putname(nd->last.name); |
1da177e4 LT |
1856 | goto exit; |
1857 | } | |
4ac91378 | 1858 | dir = nd->path.dentry; |
1b1dcc1b | 1859 | mutex_lock(&dir->d_inode->i_mutex); |
49705b77 | 1860 | path.dentry = lookup_hash(nd); |
4ac91378 | 1861 | path.mnt = nd->path.mnt; |
82984114 | 1862 | __putname(nd->last.name); |
1da177e4 LT |
1863 | goto do_last; |
1864 | } | |
1865 | ||
1866 | /** | |
1867 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1868 | * @nd: nameidata info | |
1869 | * @is_dir: directory flag | |
1870 | * | |
1871 | * Simple function to lookup and return a dentry and create it | |
1872 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1873 | * |
4ac91378 | 1874 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1875 | */ |
1876 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1877 | { | |
c663e5d8 | 1878 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1879 | |
4ac91378 | 1880 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1881 | /* |
1882 | * Yucky last component or no last component at all? | |
1883 | * (foo/., foo/.., /////) | |
1884 | */ | |
1da177e4 LT |
1885 | if (nd->last_type != LAST_NORM) |
1886 | goto fail; | |
1887 | nd->flags &= ~LOOKUP_PARENT; | |
a634904a AM |
1888 | nd->flags |= LOOKUP_CREATE; |
1889 | nd->intent.open.flags = O_EXCL; | |
c663e5d8 CH |
1890 | |
1891 | /* | |
1892 | * Do the final lookup. | |
1893 | */ | |
49705b77 | 1894 | dentry = lookup_hash(nd); |
1da177e4 LT |
1895 | if (IS_ERR(dentry)) |
1896 | goto fail; | |
c663e5d8 CH |
1897 | |
1898 | /* | |
1899 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1900 | * From the vfs_mknod() POV we just have a negative dentry - | |
1901 | * all is fine. Let's be bastards - you had / on the end, you've | |
1902 | * been asking for (non-existent) directory. -ENOENT for you. | |
1903 | */ | |
1da177e4 LT |
1904 | if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode) |
1905 | goto enoent; | |
1906 | return dentry; | |
1907 | enoent: | |
1908 | dput(dentry); | |
1909 | dentry = ERR_PTR(-ENOENT); | |
1910 | fail: | |
1911 | return dentry; | |
1912 | } | |
f81a0bff | 1913 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1914 | |
1915 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1916 | { | |
1917 | int error = may_create(dir, dentry, NULL); | |
1918 | ||
1919 | if (error) | |
1920 | return error; | |
1921 | ||
1922 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1923 | return -EPERM; | |
1924 | ||
1925 | if (!dir->i_op || !dir->i_op->mknod) | |
1926 | return -EPERM; | |
1927 | ||
1928 | error = security_inode_mknod(dir, dentry, mode, dev); | |
1929 | if (error) | |
1930 | return error; | |
1931 | ||
1932 | DQUOT_INIT(dir); | |
1933 | error = dir->i_op->mknod(dir, dentry, mode, dev); | |
a74574aa | 1934 | if (!error) |
f38aa942 | 1935 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1936 | return error; |
1937 | } | |
1938 | ||
5590ff0d UD |
1939 | asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode, |
1940 | unsigned dev) | |
1da177e4 LT |
1941 | { |
1942 | int error = 0; | |
1943 | char * tmp; | |
1944 | struct dentry * dentry; | |
1945 | struct nameidata nd; | |
1946 | ||
1947 | if (S_ISDIR(mode)) | |
1948 | return -EPERM; | |
1949 | tmp = getname(filename); | |
1950 | if (IS_ERR(tmp)) | |
1951 | return PTR_ERR(tmp); | |
1952 | ||
5590ff0d | 1953 | error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd); |
1da177e4 LT |
1954 | if (error) |
1955 | goto out; | |
1956 | dentry = lookup_create(&nd, 0); | |
1957 | error = PTR_ERR(dentry); | |
1958 | ||
4ac91378 | 1959 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
1da177e4 LT |
1960 | mode &= ~current->fs->umask; |
1961 | if (!IS_ERR(dentry)) { | |
1962 | switch (mode & S_IFMT) { | |
1963 | case 0: case S_IFREG: | |
4ac91378 | 1964 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
1965 | break; |
1966 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 1967 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
1968 | new_decode_dev(dev)); |
1969 | break; | |
1970 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 1971 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 LT |
1972 | break; |
1973 | case S_IFDIR: | |
1974 | error = -EPERM; | |
1975 | break; | |
1976 | default: | |
1977 | error = -EINVAL; | |
1978 | } | |
1979 | dput(dentry); | |
1980 | } | |
4ac91378 | 1981 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
1982 | path_release(&nd); |
1983 | out: | |
1984 | putname(tmp); | |
1985 | ||
1986 | return error; | |
1987 | } | |
1988 | ||
5590ff0d UD |
1989 | asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev) |
1990 | { | |
1991 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
1992 | } | |
1993 | ||
1da177e4 LT |
1994 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
1995 | { | |
1996 | int error = may_create(dir, dentry, NULL); | |
1997 | ||
1998 | if (error) | |
1999 | return error; | |
2000 | ||
2001 | if (!dir->i_op || !dir->i_op->mkdir) | |
2002 | return -EPERM; | |
2003 | ||
2004 | mode &= (S_IRWXUGO|S_ISVTX); | |
2005 | error = security_inode_mkdir(dir, dentry, mode); | |
2006 | if (error) | |
2007 | return error; | |
2008 | ||
2009 | DQUOT_INIT(dir); | |
2010 | error = dir->i_op->mkdir(dir, dentry, mode); | |
a74574aa | 2011 | if (!error) |
f38aa942 | 2012 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2013 | return error; |
2014 | } | |
2015 | ||
5590ff0d | 2016 | asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode) |
1da177e4 LT |
2017 | { |
2018 | int error = 0; | |
2019 | char * tmp; | |
6902d925 DH |
2020 | struct dentry *dentry; |
2021 | struct nameidata nd; | |
1da177e4 LT |
2022 | |
2023 | tmp = getname(pathname); | |
2024 | error = PTR_ERR(tmp); | |
6902d925 DH |
2025 | if (IS_ERR(tmp)) |
2026 | goto out_err; | |
1da177e4 | 2027 | |
6902d925 DH |
2028 | error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd); |
2029 | if (error) | |
2030 | goto out; | |
2031 | dentry = lookup_create(&nd, 1); | |
2032 | error = PTR_ERR(dentry); | |
2033 | if (IS_ERR(dentry)) | |
2034 | goto out_unlock; | |
1da177e4 | 2035 | |
4ac91378 | 2036 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
6902d925 | 2037 | mode &= ~current->fs->umask; |
4ac91378 | 2038 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
6902d925 DH |
2039 | dput(dentry); |
2040 | out_unlock: | |
4ac91378 | 2041 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
6902d925 DH |
2042 | path_release(&nd); |
2043 | out: | |
2044 | putname(tmp); | |
2045 | out_err: | |
1da177e4 LT |
2046 | return error; |
2047 | } | |
2048 | ||
5590ff0d UD |
2049 | asmlinkage long sys_mkdir(const char __user *pathname, int mode) |
2050 | { | |
2051 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2052 | } | |
2053 | ||
1da177e4 LT |
2054 | /* |
2055 | * We try to drop the dentry early: we should have | |
2056 | * a usage count of 2 if we're the only user of this | |
2057 | * dentry, and if that is true (possibly after pruning | |
2058 | * the dcache), then we drop the dentry now. | |
2059 | * | |
2060 | * A low-level filesystem can, if it choses, legally | |
2061 | * do a | |
2062 | * | |
2063 | * if (!d_unhashed(dentry)) | |
2064 | * return -EBUSY; | |
2065 | * | |
2066 | * if it cannot handle the case of removing a directory | |
2067 | * that is still in use by something else.. | |
2068 | */ | |
2069 | void dentry_unhash(struct dentry *dentry) | |
2070 | { | |
2071 | dget(dentry); | |
dc168427 | 2072 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2073 | spin_lock(&dcache_lock); |
2074 | spin_lock(&dentry->d_lock); | |
2075 | if (atomic_read(&dentry->d_count) == 2) | |
2076 | __d_drop(dentry); | |
2077 | spin_unlock(&dentry->d_lock); | |
2078 | spin_unlock(&dcache_lock); | |
2079 | } | |
2080 | ||
2081 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2082 | { | |
2083 | int error = may_delete(dir, dentry, 1); | |
2084 | ||
2085 | if (error) | |
2086 | return error; | |
2087 | ||
2088 | if (!dir->i_op || !dir->i_op->rmdir) | |
2089 | return -EPERM; | |
2090 | ||
2091 | DQUOT_INIT(dir); | |
2092 | ||
1b1dcc1b | 2093 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2094 | dentry_unhash(dentry); |
2095 | if (d_mountpoint(dentry)) | |
2096 | error = -EBUSY; | |
2097 | else { | |
2098 | error = security_inode_rmdir(dir, dentry); | |
2099 | if (!error) { | |
2100 | error = dir->i_op->rmdir(dir, dentry); | |
2101 | if (!error) | |
2102 | dentry->d_inode->i_flags |= S_DEAD; | |
2103 | } | |
2104 | } | |
1b1dcc1b | 2105 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2106 | if (!error) { |
1da177e4 LT |
2107 | d_delete(dentry); |
2108 | } | |
2109 | dput(dentry); | |
2110 | ||
2111 | return error; | |
2112 | } | |
2113 | ||
5590ff0d | 2114 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2115 | { |
2116 | int error = 0; | |
2117 | char * name; | |
2118 | struct dentry *dentry; | |
2119 | struct nameidata nd; | |
2120 | ||
2121 | name = getname(pathname); | |
2122 | if(IS_ERR(name)) | |
2123 | return PTR_ERR(name); | |
2124 | ||
5590ff0d | 2125 | error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2126 | if (error) |
2127 | goto exit; | |
2128 | ||
2129 | switch(nd.last_type) { | |
2130 | case LAST_DOTDOT: | |
2131 | error = -ENOTEMPTY; | |
2132 | goto exit1; | |
2133 | case LAST_DOT: | |
2134 | error = -EINVAL; | |
2135 | goto exit1; | |
2136 | case LAST_ROOT: | |
2137 | error = -EBUSY; | |
2138 | goto exit1; | |
2139 | } | |
4ac91378 | 2140 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2141 | dentry = lookup_hash(&nd); |
1da177e4 | 2142 | error = PTR_ERR(dentry); |
6902d925 DH |
2143 | if (IS_ERR(dentry)) |
2144 | goto exit2; | |
4ac91378 | 2145 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
6902d925 DH |
2146 | dput(dentry); |
2147 | exit2: | |
4ac91378 | 2148 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2149 | exit1: |
2150 | path_release(&nd); | |
2151 | exit: | |
2152 | putname(name); | |
2153 | return error; | |
2154 | } | |
2155 | ||
5590ff0d UD |
2156 | asmlinkage long sys_rmdir(const char __user *pathname) |
2157 | { | |
2158 | return do_rmdir(AT_FDCWD, pathname); | |
2159 | } | |
2160 | ||
1da177e4 LT |
2161 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2162 | { | |
2163 | int error = may_delete(dir, dentry, 0); | |
2164 | ||
2165 | if (error) | |
2166 | return error; | |
2167 | ||
2168 | if (!dir->i_op || !dir->i_op->unlink) | |
2169 | return -EPERM; | |
2170 | ||
2171 | DQUOT_INIT(dir); | |
2172 | ||
1b1dcc1b | 2173 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2174 | if (d_mountpoint(dentry)) |
2175 | error = -EBUSY; | |
2176 | else { | |
2177 | error = security_inode_unlink(dir, dentry); | |
2178 | if (!error) | |
2179 | error = dir->i_op->unlink(dir, dentry); | |
2180 | } | |
1b1dcc1b | 2181 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2182 | |
2183 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2184 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2185 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2186 | d_delete(dentry); |
1da177e4 | 2187 | } |
0eeca283 | 2188 | |
1da177e4 LT |
2189 | return error; |
2190 | } | |
2191 | ||
2192 | /* | |
2193 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2194 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2195 | * writeout happening, and we don't want to prevent access to the directory |
2196 | * while waiting on the I/O. | |
2197 | */ | |
5590ff0d | 2198 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 LT |
2199 | { |
2200 | int error = 0; | |
2201 | char * name; | |
2202 | struct dentry *dentry; | |
2203 | struct nameidata nd; | |
2204 | struct inode *inode = NULL; | |
2205 | ||
2206 | name = getname(pathname); | |
2207 | if(IS_ERR(name)) | |
2208 | return PTR_ERR(name); | |
2209 | ||
5590ff0d | 2210 | error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2211 | if (error) |
2212 | goto exit; | |
2213 | error = -EISDIR; | |
2214 | if (nd.last_type != LAST_NORM) | |
2215 | goto exit1; | |
4ac91378 | 2216 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2217 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2218 | error = PTR_ERR(dentry); |
2219 | if (!IS_ERR(dentry)) { | |
2220 | /* Why not before? Because we want correct error value */ | |
2221 | if (nd.last.name[nd.last.len]) | |
2222 | goto slashes; | |
2223 | inode = dentry->d_inode; | |
2224 | if (inode) | |
2225 | atomic_inc(&inode->i_count); | |
4ac91378 | 2226 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
1da177e4 LT |
2227 | exit2: |
2228 | dput(dentry); | |
2229 | } | |
4ac91378 | 2230 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2231 | if (inode) |
2232 | iput(inode); /* truncate the inode here */ | |
2233 | exit1: | |
2234 | path_release(&nd); | |
2235 | exit: | |
2236 | putname(name); | |
2237 | return error; | |
2238 | ||
2239 | slashes: | |
2240 | error = !dentry->d_inode ? -ENOENT : | |
2241 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2242 | goto exit2; | |
2243 | } | |
2244 | ||
5590ff0d UD |
2245 | asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag) |
2246 | { | |
2247 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2248 | return -EINVAL; | |
2249 | ||
2250 | if (flag & AT_REMOVEDIR) | |
2251 | return do_rmdir(dfd, pathname); | |
2252 | ||
2253 | return do_unlinkat(dfd, pathname); | |
2254 | } | |
2255 | ||
2256 | asmlinkage long sys_unlink(const char __user *pathname) | |
2257 | { | |
2258 | return do_unlinkat(AT_FDCWD, pathname); | |
2259 | } | |
2260 | ||
1da177e4 LT |
2261 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname, int mode) |
2262 | { | |
2263 | int error = may_create(dir, dentry, NULL); | |
2264 | ||
2265 | if (error) | |
2266 | return error; | |
2267 | ||
2268 | if (!dir->i_op || !dir->i_op->symlink) | |
2269 | return -EPERM; | |
2270 | ||
2271 | error = security_inode_symlink(dir, dentry, oldname); | |
2272 | if (error) | |
2273 | return error; | |
2274 | ||
2275 | DQUOT_INIT(dir); | |
2276 | error = dir->i_op->symlink(dir, dentry, oldname); | |
a74574aa | 2277 | if (!error) |
f38aa942 | 2278 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2279 | return error; |
2280 | } | |
2281 | ||
5590ff0d UD |
2282 | asmlinkage long sys_symlinkat(const char __user *oldname, |
2283 | int newdfd, const char __user *newname) | |
1da177e4 LT |
2284 | { |
2285 | int error = 0; | |
2286 | char * from; | |
2287 | char * to; | |
6902d925 DH |
2288 | struct dentry *dentry; |
2289 | struct nameidata nd; | |
1da177e4 LT |
2290 | |
2291 | from = getname(oldname); | |
2292 | if(IS_ERR(from)) | |
2293 | return PTR_ERR(from); | |
2294 | to = getname(newname); | |
2295 | error = PTR_ERR(to); | |
6902d925 DH |
2296 | if (IS_ERR(to)) |
2297 | goto out_putname; | |
1da177e4 | 2298 | |
6902d925 DH |
2299 | error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd); |
2300 | if (error) | |
2301 | goto out; | |
2302 | dentry = lookup_create(&nd, 0); | |
2303 | error = PTR_ERR(dentry); | |
2304 | if (IS_ERR(dentry)) | |
2305 | goto out_unlock; | |
2306 | ||
4ac91378 | 2307 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from, S_IALLUGO); |
6902d925 DH |
2308 | dput(dentry); |
2309 | out_unlock: | |
4ac91378 | 2310 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
6902d925 | 2311 | path_release(&nd); |
1da177e4 | 2312 | out: |
6902d925 DH |
2313 | putname(to); |
2314 | out_putname: | |
1da177e4 LT |
2315 | putname(from); |
2316 | return error; | |
2317 | } | |
2318 | ||
5590ff0d UD |
2319 | asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname) |
2320 | { | |
2321 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2322 | } | |
2323 | ||
1da177e4 LT |
2324 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2325 | { | |
2326 | struct inode *inode = old_dentry->d_inode; | |
2327 | int error; | |
2328 | ||
2329 | if (!inode) | |
2330 | return -ENOENT; | |
2331 | ||
2332 | error = may_create(dir, new_dentry, NULL); | |
2333 | if (error) | |
2334 | return error; | |
2335 | ||
2336 | if (dir->i_sb != inode->i_sb) | |
2337 | return -EXDEV; | |
2338 | ||
2339 | /* | |
2340 | * A link to an append-only or immutable file cannot be created. | |
2341 | */ | |
2342 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2343 | return -EPERM; | |
2344 | if (!dir->i_op || !dir->i_op->link) | |
2345 | return -EPERM; | |
2346 | if (S_ISDIR(old_dentry->d_inode->i_mode)) | |
2347 | return -EPERM; | |
2348 | ||
2349 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2350 | if (error) | |
2351 | return error; | |
2352 | ||
1b1dcc1b | 2353 | mutex_lock(&old_dentry->d_inode->i_mutex); |
1da177e4 LT |
2354 | DQUOT_INIT(dir); |
2355 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
1b1dcc1b | 2356 | mutex_unlock(&old_dentry->d_inode->i_mutex); |
e31e14ec | 2357 | if (!error) |
ece95912 | 2358 | fsnotify_link(dir, old_dentry->d_inode, new_dentry); |
1da177e4 LT |
2359 | return error; |
2360 | } | |
2361 | ||
2362 | /* | |
2363 | * Hardlinks are often used in delicate situations. We avoid | |
2364 | * security-related surprises by not following symlinks on the | |
2365 | * newname. --KAB | |
2366 | * | |
2367 | * We don't follow them on the oldname either to be compatible | |
2368 | * with linux 2.0, and to avoid hard-linking to directories | |
2369 | * and other special files. --ADM | |
2370 | */ | |
5590ff0d | 2371 | asmlinkage long sys_linkat(int olddfd, const char __user *oldname, |
c04030e1 UD |
2372 | int newdfd, const char __user *newname, |
2373 | int flags) | |
1da177e4 LT |
2374 | { |
2375 | struct dentry *new_dentry; | |
2376 | struct nameidata nd, old_nd; | |
2377 | int error; | |
2378 | char * to; | |
2379 | ||
45c9b11a | 2380 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2381 | return -EINVAL; |
2382 | ||
1da177e4 LT |
2383 | to = getname(newname); |
2384 | if (IS_ERR(to)) | |
2385 | return PTR_ERR(to); | |
2386 | ||
45c9b11a UD |
2387 | error = __user_walk_fd(olddfd, oldname, |
2388 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2389 | &old_nd); | |
1da177e4 LT |
2390 | if (error) |
2391 | goto exit; | |
5590ff0d | 2392 | error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2393 | if (error) |
2394 | goto out; | |
2395 | error = -EXDEV; | |
4ac91378 | 2396 | if (old_nd.path.mnt != nd.path.mnt) |
1da177e4 LT |
2397 | goto out_release; |
2398 | new_dentry = lookup_create(&nd, 0); | |
2399 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2400 | if (IS_ERR(new_dentry)) |
2401 | goto out_unlock; | |
4ac91378 | 2402 | error = vfs_link(old_nd.path.dentry, nd.path.dentry->d_inode, new_dentry); |
6902d925 DH |
2403 | dput(new_dentry); |
2404 | out_unlock: | |
4ac91378 | 2405 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2406 | out_release: |
2407 | path_release(&nd); | |
2408 | out: | |
2409 | path_release(&old_nd); | |
2410 | exit: | |
2411 | putname(to); | |
2412 | ||
2413 | return error; | |
2414 | } | |
2415 | ||
5590ff0d UD |
2416 | asmlinkage long sys_link(const char __user *oldname, const char __user *newname) |
2417 | { | |
c04030e1 | 2418 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2419 | } |
2420 | ||
1da177e4 LT |
2421 | /* |
2422 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2423 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2424 | * Problems: | |
2425 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2426 | * b) race potential - two innocent renames can create a loop together. | |
2427 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2428 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2429 | * story. |
2430 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2431 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2432 | * whether the target exists). Solution: try to be smart with locking |
2433 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2434 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2435 | * move will be locked. Thus we can rank directories by the tree |
2436 | * (ancestors first) and rank all non-directories after them. | |
2437 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2438 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2439 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2440 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2441 | * we'd better make sure that there's no link(2) for them. | |
2442 | * d) some filesystems don't support opened-but-unlinked directories, | |
2443 | * either because of layout or because they are not ready to deal with | |
2444 | * all cases correctly. The latter will be fixed (taking this sort of | |
2445 | * stuff into VFS), but the former is not going away. Solution: the same | |
2446 | * trick as in rmdir(). | |
2447 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2448 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2449 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2450 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2451 | * locking]. |
2452 | */ | |
75c96f85 AB |
2453 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2454 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2455 | { |
2456 | int error = 0; | |
2457 | struct inode *target; | |
2458 | ||
2459 | /* | |
2460 | * If we are going to change the parent - check write permissions, | |
2461 | * we'll need to flip '..'. | |
2462 | */ | |
2463 | if (new_dir != old_dir) { | |
2464 | error = permission(old_dentry->d_inode, MAY_WRITE, NULL); | |
2465 | if (error) | |
2466 | return error; | |
2467 | } | |
2468 | ||
2469 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2470 | if (error) | |
2471 | return error; | |
2472 | ||
2473 | target = new_dentry->d_inode; | |
2474 | if (target) { | |
1b1dcc1b | 2475 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2476 | dentry_unhash(new_dentry); |
2477 | } | |
2478 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2479 | error = -EBUSY; | |
2480 | else | |
2481 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2482 | if (target) { | |
2483 | if (!error) | |
2484 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2485 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2486 | if (d_unhashed(new_dentry)) |
2487 | d_rehash(new_dentry); | |
2488 | dput(new_dentry); | |
2489 | } | |
e31e14ec | 2490 | if (!error) |
349457cc MF |
2491 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2492 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2493 | return error; |
2494 | } | |
2495 | ||
75c96f85 AB |
2496 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2497 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2498 | { |
2499 | struct inode *target; | |
2500 | int error; | |
2501 | ||
2502 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2503 | if (error) | |
2504 | return error; | |
2505 | ||
2506 | dget(new_dentry); | |
2507 | target = new_dentry->d_inode; | |
2508 | if (target) | |
1b1dcc1b | 2509 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2510 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2511 | error = -EBUSY; | |
2512 | else | |
2513 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2514 | if (!error) { | |
349457cc | 2515 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2516 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2517 | } |
2518 | if (target) | |
1b1dcc1b | 2519 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2520 | dput(new_dentry); |
2521 | return error; | |
2522 | } | |
2523 | ||
2524 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2525 | struct inode *new_dir, struct dentry *new_dentry) | |
2526 | { | |
2527 | int error; | |
2528 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2529 | const char *old_name; |
1da177e4 LT |
2530 | |
2531 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2532 | return 0; | |
2533 | ||
2534 | error = may_delete(old_dir, old_dentry, is_dir); | |
2535 | if (error) | |
2536 | return error; | |
2537 | ||
2538 | if (!new_dentry->d_inode) | |
2539 | error = may_create(new_dir, new_dentry, NULL); | |
2540 | else | |
2541 | error = may_delete(new_dir, new_dentry, is_dir); | |
2542 | if (error) | |
2543 | return error; | |
2544 | ||
2545 | if (!old_dir->i_op || !old_dir->i_op->rename) | |
2546 | return -EPERM; | |
2547 | ||
2548 | DQUOT_INIT(old_dir); | |
2549 | DQUOT_INIT(new_dir); | |
2550 | ||
0eeca283 RL |
2551 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2552 | ||
1da177e4 LT |
2553 | if (is_dir) |
2554 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2555 | else | |
2556 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2557 | if (!error) { | |
0eeca283 | 2558 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2559 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2560 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2561 | } |
0eeca283 RL |
2562 | fsnotify_oldname_free(old_name); |
2563 | ||
1da177e4 LT |
2564 | return error; |
2565 | } | |
2566 | ||
5590ff0d UD |
2567 | static int do_rename(int olddfd, const char *oldname, |
2568 | int newdfd, const char *newname) | |
1da177e4 LT |
2569 | { |
2570 | int error = 0; | |
2571 | struct dentry * old_dir, * new_dir; | |
2572 | struct dentry * old_dentry, *new_dentry; | |
2573 | struct dentry * trap; | |
2574 | struct nameidata oldnd, newnd; | |
2575 | ||
5590ff0d | 2576 | error = do_path_lookup(olddfd, oldname, LOOKUP_PARENT, &oldnd); |
1da177e4 LT |
2577 | if (error) |
2578 | goto exit; | |
2579 | ||
5590ff0d | 2580 | error = do_path_lookup(newdfd, newname, LOOKUP_PARENT, &newnd); |
1da177e4 LT |
2581 | if (error) |
2582 | goto exit1; | |
2583 | ||
2584 | error = -EXDEV; | |
4ac91378 | 2585 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2586 | goto exit2; |
2587 | ||
4ac91378 | 2588 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2589 | error = -EBUSY; |
2590 | if (oldnd.last_type != LAST_NORM) | |
2591 | goto exit2; | |
2592 | ||
4ac91378 | 2593 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2594 | if (newnd.last_type != LAST_NORM) |
2595 | goto exit2; | |
2596 | ||
2597 | trap = lock_rename(new_dir, old_dir); | |
2598 | ||
49705b77 | 2599 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2600 | error = PTR_ERR(old_dentry); |
2601 | if (IS_ERR(old_dentry)) | |
2602 | goto exit3; | |
2603 | /* source must exist */ | |
2604 | error = -ENOENT; | |
2605 | if (!old_dentry->d_inode) | |
2606 | goto exit4; | |
2607 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2608 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2609 | error = -ENOTDIR; | |
2610 | if (oldnd.last.name[oldnd.last.len]) | |
2611 | goto exit4; | |
2612 | if (newnd.last.name[newnd.last.len]) | |
2613 | goto exit4; | |
2614 | } | |
2615 | /* source should not be ancestor of target */ | |
2616 | error = -EINVAL; | |
2617 | if (old_dentry == trap) | |
2618 | goto exit4; | |
49705b77 | 2619 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2620 | error = PTR_ERR(new_dentry); |
2621 | if (IS_ERR(new_dentry)) | |
2622 | goto exit4; | |
2623 | /* target should not be an ancestor of source */ | |
2624 | error = -ENOTEMPTY; | |
2625 | if (new_dentry == trap) | |
2626 | goto exit5; | |
2627 | ||
2628 | error = vfs_rename(old_dir->d_inode, old_dentry, | |
2629 | new_dir->d_inode, new_dentry); | |
2630 | exit5: | |
2631 | dput(new_dentry); | |
2632 | exit4: | |
2633 | dput(old_dentry); | |
2634 | exit3: | |
2635 | unlock_rename(new_dir, old_dir); | |
2636 | exit2: | |
2637 | path_release(&newnd); | |
2638 | exit1: | |
2639 | path_release(&oldnd); | |
2640 | exit: | |
2641 | return error; | |
2642 | } | |
2643 | ||
5590ff0d UD |
2644 | asmlinkage long sys_renameat(int olddfd, const char __user *oldname, |
2645 | int newdfd, const char __user *newname) | |
1da177e4 LT |
2646 | { |
2647 | int error; | |
2648 | char * from; | |
2649 | char * to; | |
2650 | ||
2651 | from = getname(oldname); | |
2652 | if(IS_ERR(from)) | |
2653 | return PTR_ERR(from); | |
2654 | to = getname(newname); | |
2655 | error = PTR_ERR(to); | |
2656 | if (!IS_ERR(to)) { | |
5590ff0d | 2657 | error = do_rename(olddfd, from, newdfd, to); |
1da177e4 LT |
2658 | putname(to); |
2659 | } | |
2660 | putname(from); | |
2661 | return error; | |
2662 | } | |
2663 | ||
5590ff0d UD |
2664 | asmlinkage long sys_rename(const char __user *oldname, const char __user *newname) |
2665 | { | |
2666 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2667 | } | |
2668 | ||
1da177e4 LT |
2669 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2670 | { | |
2671 | int len; | |
2672 | ||
2673 | len = PTR_ERR(link); | |
2674 | if (IS_ERR(link)) | |
2675 | goto out; | |
2676 | ||
2677 | len = strlen(link); | |
2678 | if (len > (unsigned) buflen) | |
2679 | len = buflen; | |
2680 | if (copy_to_user(buffer, link, len)) | |
2681 | len = -EFAULT; | |
2682 | out: | |
2683 | return len; | |
2684 | } | |
2685 | ||
2686 | /* | |
2687 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2688 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2689 | * using) it for any given inode is up to filesystem. | |
2690 | */ | |
2691 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2692 | { | |
2693 | struct nameidata nd; | |
cc314eef LT |
2694 | void *cookie; |
2695 | ||
1da177e4 | 2696 | nd.depth = 0; |
cc314eef LT |
2697 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
2698 | if (!IS_ERR(cookie)) { | |
2699 | int res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
1da177e4 | 2700 | if (dentry->d_inode->i_op->put_link) |
cc314eef LT |
2701 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); |
2702 | cookie = ERR_PTR(res); | |
1da177e4 | 2703 | } |
cc314eef | 2704 | return PTR_ERR(cookie); |
1da177e4 LT |
2705 | } |
2706 | ||
2707 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2708 | { | |
2709 | return __vfs_follow_link(nd, link); | |
2710 | } | |
2711 | ||
2712 | /* get the link contents into pagecache */ | |
2713 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2714 | { | |
2715 | struct page * page; | |
2716 | struct address_space *mapping = dentry->d_inode->i_mapping; | |
090d2b18 | 2717 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2718 | if (IS_ERR(page)) |
6fe6900e | 2719 | return (char*)page; |
1da177e4 LT |
2720 | *ppage = page; |
2721 | return kmap(page); | |
1da177e4 LT |
2722 | } |
2723 | ||
2724 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2725 | { | |
2726 | struct page *page = NULL; | |
2727 | char *s = page_getlink(dentry, &page); | |
2728 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2729 | if (page) { | |
2730 | kunmap(page); | |
2731 | page_cache_release(page); | |
2732 | } | |
2733 | return res; | |
2734 | } | |
2735 | ||
cc314eef | 2736 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2737 | { |
cc314eef | 2738 | struct page *page = NULL; |
1da177e4 | 2739 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2740 | return page; |
1da177e4 LT |
2741 | } |
2742 | ||
cc314eef | 2743 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2744 | { |
cc314eef LT |
2745 | struct page *page = cookie; |
2746 | ||
2747 | if (page) { | |
1da177e4 LT |
2748 | kunmap(page); |
2749 | page_cache_release(page); | |
1da177e4 LT |
2750 | } |
2751 | } | |
2752 | ||
0adb25d2 KK |
2753 | int __page_symlink(struct inode *inode, const char *symname, int len, |
2754 | gfp_t gfp_mask) | |
1da177e4 LT |
2755 | { |
2756 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2757 | struct page *page; |
afddba49 | 2758 | void *fsdata; |
beb497ab | 2759 | int err; |
1da177e4 LT |
2760 | char *kaddr; |
2761 | ||
7e53cac4 | 2762 | retry: |
afddba49 NP |
2763 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
2764 | AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata); | |
1da177e4 | 2765 | if (err) |
afddba49 NP |
2766 | goto fail; |
2767 | ||
1da177e4 LT |
2768 | kaddr = kmap_atomic(page, KM_USER0); |
2769 | memcpy(kaddr, symname, len-1); | |
2770 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2771 | |
2772 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2773 | page, fsdata); | |
1da177e4 LT |
2774 | if (err < 0) |
2775 | goto fail; | |
afddba49 NP |
2776 | if (err < len-1) |
2777 | goto retry; | |
2778 | ||
1da177e4 LT |
2779 | mark_inode_dirty(inode); |
2780 | return 0; | |
1da177e4 LT |
2781 | fail: |
2782 | return err; | |
2783 | } | |
2784 | ||
0adb25d2 KK |
2785 | int page_symlink(struct inode *inode, const char *symname, int len) |
2786 | { | |
2787 | return __page_symlink(inode, symname, len, | |
2788 | mapping_gfp_mask(inode->i_mapping)); | |
2789 | } | |
2790 | ||
92e1d5be | 2791 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2792 | .readlink = generic_readlink, |
2793 | .follow_link = page_follow_link_light, | |
2794 | .put_link = page_put_link, | |
2795 | }; | |
2796 | ||
2797 | EXPORT_SYMBOL(__user_walk); | |
5590ff0d | 2798 | EXPORT_SYMBOL(__user_walk_fd); |
1da177e4 LT |
2799 | EXPORT_SYMBOL(follow_down); |
2800 | EXPORT_SYMBOL(follow_up); | |
2801 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2802 | EXPORT_SYMBOL(getname); | |
2803 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2804 | EXPORT_SYMBOL(lookup_one_len); |
2805 | EXPORT_SYMBOL(page_follow_link_light); | |
2806 | EXPORT_SYMBOL(page_put_link); | |
2807 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2808 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2809 | EXPORT_SYMBOL(page_symlink); |
2810 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2811 | EXPORT_SYMBOL(path_lookup); | |
16f18200 | 2812 | EXPORT_SYMBOL(vfs_path_lookup); |
1da177e4 | 2813 | EXPORT_SYMBOL(path_release); |
1da177e4 | 2814 | EXPORT_SYMBOL(permission); |
e4543edd | 2815 | EXPORT_SYMBOL(vfs_permission); |
8c744fb8 | 2816 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2817 | EXPORT_SYMBOL(unlock_rename); |
2818 | EXPORT_SYMBOL(vfs_create); | |
2819 | EXPORT_SYMBOL(vfs_follow_link); | |
2820 | EXPORT_SYMBOL(vfs_link); | |
2821 | EXPORT_SYMBOL(vfs_mkdir); | |
2822 | EXPORT_SYMBOL(vfs_mknod); | |
2823 | EXPORT_SYMBOL(generic_permission); | |
2824 | EXPORT_SYMBOL(vfs_readlink); | |
2825 | EXPORT_SYMBOL(vfs_rename); | |
2826 | EXPORT_SYMBOL(vfs_rmdir); | |
2827 | EXPORT_SYMBOL(vfs_symlink); | |
2828 | EXPORT_SYMBOL(vfs_unlink); | |
2829 | EXPORT_SYMBOL(dentry_unhash); | |
2830 | EXPORT_SYMBOL(generic_readlink); |