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/smp_lock.h> |
26 | #include <linux/personality.h> | |
27 | #include <linux/security.h> | |
28 | #include <linux/syscalls.h> | |
29 | #include <linux/mount.h> | |
30 | #include <linux/audit.h> | |
16f7e0fe | 31 | #include <linux/capability.h> |
834f2a4a | 32 | #include <linux/file.h> |
5590ff0d UD |
33 | #include <linux/fcntl.h> |
34 | #include <linux/namei.h> | |
1da177e4 LT |
35 | #include <asm/namei.h> |
36 | #include <asm/uaccess.h> | |
37 | ||
38 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
39 | ||
40 | /* [Feb-1997 T. Schoebel-Theuer] | |
41 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
42 | * were necessary because of omirr. The reason is that omirr needs | |
43 | * to know the _real_ pathname, not the user-supplied one, in case | |
44 | * of symlinks (and also when transname replacements occur). | |
45 | * | |
46 | * The new code replaces the old recursive symlink resolution with | |
47 | * an iterative one (in case of non-nested symlink chains). It does | |
48 | * this with calls to <fs>_follow_link(). | |
49 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
50 | * replaced with a single function lookup_dentry() that can handle all | |
51 | * the special cases of the former code. | |
52 | * | |
53 | * With the new dcache, the pathname is stored at each inode, at least as | |
54 | * long as the refcount of the inode is positive. As a side effect, the | |
55 | * size of the dcache depends on the inode cache and thus is dynamic. | |
56 | * | |
57 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
58 | * resolution to correspond with current state of the code. | |
59 | * | |
60 | * Note that the symlink resolution is not *completely* iterative. | |
61 | * There is still a significant amount of tail- and mid- recursion in | |
62 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
63 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
64 | * may return different results than <fs>_follow_link(). Many virtual | |
65 | * filesystems (including /proc) exhibit this behavior. | |
66 | */ | |
67 | ||
68 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
69 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
70 | * and the name already exists in form of a symlink, try to create the new | |
71 | * name indicated by the symlink. The old code always complained that the | |
72 | * name already exists, due to not following the symlink even if its target | |
73 | * is nonexistent. The new semantics affects also mknod() and link() when | |
74 | * the name is a symlink pointing to a non-existant name. | |
75 | * | |
76 | * I don't know which semantics is the right one, since I have no access | |
77 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
78 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
79 | * "old" one. Personally, I think the new semantics is much more logical. | |
80 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
81 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
82 | * and in the old Linux semantics. | |
83 | */ | |
84 | ||
85 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
86 | * semantics. See the comments in "open_namei" and "do_link" below. | |
87 | * | |
88 | * [10-Sep-98 Alan Modra] Another symlink change. | |
89 | */ | |
90 | ||
91 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
92 | * inside the path - always follow. | |
93 | * in the last component in creation/removal/renaming - never follow. | |
94 | * if LOOKUP_FOLLOW passed - follow. | |
95 | * if the pathname has trailing slashes - follow. | |
96 | * otherwise - don't follow. | |
97 | * (applied in that order). | |
98 | * | |
99 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
100 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
101 | * During the 2.4 we need to fix the userland stuff depending on it - | |
102 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
103 | * XEmacs seems to be relying on it... | |
104 | */ | |
105 | /* | |
106 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 107 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
108 | * any extra contention... |
109 | */ | |
110 | ||
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 | { | |
a343bb77 | 230 | umode_t mode = inode->i_mode; |
1da177e4 LT |
231 | int retval, submask; |
232 | ||
233 | if (mask & MAY_WRITE) { | |
1da177e4 LT |
234 | |
235 | /* | |
236 | * Nobody gets write access to a read-only fs. | |
237 | */ | |
238 | if (IS_RDONLY(inode) && | |
239 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
240 | return -EROFS; | |
241 | ||
242 | /* | |
243 | * Nobody gets write access to an immutable file. | |
244 | */ | |
245 | if (IS_IMMUTABLE(inode)) | |
246 | return -EACCES; | |
247 | } | |
248 | ||
249 | ||
a343bb77 TM |
250 | /* |
251 | * MAY_EXEC on regular files requires special handling: We override | |
317a40ac SS |
252 | * filesystem execute permissions if the mode bits aren't set or |
253 | * the fs is mounted with the "noexec" flag. | |
a343bb77 | 254 | */ |
317a40ac SS |
255 | if ((mask & MAY_EXEC) && S_ISREG(mode) && (!(mode & S_IXUGO) || |
256 | (nd && nd->mnt && (nd->mnt->mnt_flags & MNT_NOEXEC)))) | |
a343bb77 TM |
257 | return -EACCES; |
258 | ||
1da177e4 LT |
259 | /* Ordinary permission routines do not understand MAY_APPEND. */ |
260 | submask = mask & ~MAY_APPEND; | |
261 | if (inode->i_op && inode->i_op->permission) | |
262 | retval = inode->i_op->permission(inode, submask, nd); | |
263 | else | |
264 | retval = generic_permission(inode, submask, NULL); | |
265 | if (retval) | |
266 | return retval; | |
267 | ||
268 | return security_inode_permission(inode, mask, nd); | |
269 | } | |
270 | ||
e4543edd CH |
271 | /** |
272 | * vfs_permission - check for access rights to a given path | |
273 | * @nd: lookup result that describes the path | |
274 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
275 | * | |
276 | * Used to check for read/write/execute permissions on a path. | |
277 | * We use "fsuid" for this, letting us set arbitrary permissions | |
278 | * for filesystem access without changing the "normal" uids which | |
279 | * are used for other things. | |
280 | */ | |
281 | int vfs_permission(struct nameidata *nd, int mask) | |
282 | { | |
283 | return permission(nd->dentry->d_inode, mask, nd); | |
284 | } | |
285 | ||
8c744fb8 CH |
286 | /** |
287 | * file_permission - check for additional access rights to a given file | |
288 | * @file: file to check access rights for | |
289 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
290 | * | |
291 | * Used to check for read/write/execute permissions on an already opened | |
292 | * file. | |
293 | * | |
294 | * Note: | |
295 | * Do not use this function in new code. All access checks should | |
296 | * be done using vfs_permission(). | |
297 | */ | |
298 | int file_permission(struct file *file, int mask) | |
299 | { | |
0f7fc9e4 | 300 | return permission(file->f_path.dentry->d_inode, mask, NULL); |
8c744fb8 CH |
301 | } |
302 | ||
1da177e4 LT |
303 | /* |
304 | * get_write_access() gets write permission for a file. | |
305 | * put_write_access() releases this write permission. | |
306 | * This is used for regular files. | |
307 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
308 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
309 | * can have the following values: | |
310 | * 0: no writers, no VM_DENYWRITE mappings | |
311 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
312 | * > 0: (i_writecount) users are writing to the file. | |
313 | * | |
314 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
315 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
316 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
317 | * to do the change if sign is wrong. Exclusion between them is provided by | |
318 | * the inode->i_lock spinlock. | |
319 | */ | |
320 | ||
321 | int get_write_access(struct inode * inode) | |
322 | { | |
323 | spin_lock(&inode->i_lock); | |
324 | if (atomic_read(&inode->i_writecount) < 0) { | |
325 | spin_unlock(&inode->i_lock); | |
326 | return -ETXTBSY; | |
327 | } | |
328 | atomic_inc(&inode->i_writecount); | |
329 | spin_unlock(&inode->i_lock); | |
330 | ||
331 | return 0; | |
332 | } | |
333 | ||
334 | int deny_write_access(struct file * file) | |
335 | { | |
0f7fc9e4 | 336 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
337 | |
338 | spin_lock(&inode->i_lock); | |
339 | if (atomic_read(&inode->i_writecount) > 0) { | |
340 | spin_unlock(&inode->i_lock); | |
341 | return -ETXTBSY; | |
342 | } | |
343 | atomic_dec(&inode->i_writecount); | |
344 | spin_unlock(&inode->i_lock); | |
345 | ||
346 | return 0; | |
347 | } | |
348 | ||
349 | void path_release(struct nameidata *nd) | |
350 | { | |
351 | dput(nd->dentry); | |
352 | mntput(nd->mnt); | |
353 | } | |
354 | ||
355 | /* | |
356 | * umount() mustn't call path_release()/mntput() as that would clear | |
357 | * mnt_expiry_mark | |
358 | */ | |
359 | void path_release_on_umount(struct nameidata *nd) | |
360 | { | |
361 | dput(nd->dentry); | |
751c404b | 362 | mntput_no_expire(nd->mnt); |
1da177e4 LT |
363 | } |
364 | ||
834f2a4a TM |
365 | /** |
366 | * release_open_intent - free up open intent resources | |
367 | * @nd: pointer to nameidata | |
368 | */ | |
369 | void release_open_intent(struct nameidata *nd) | |
370 | { | |
0f7fc9e4 | 371 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
372 | put_filp(nd->intent.open.file); |
373 | else | |
374 | fput(nd->intent.open.file); | |
375 | } | |
376 | ||
bcdc5e01 IK |
377 | static inline struct dentry * |
378 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
379 | { | |
380 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
381 | if (unlikely(status <= 0)) { | |
382 | /* | |
383 | * The dentry failed validation. | |
384 | * If d_revalidate returned 0 attempt to invalidate | |
385 | * the dentry otherwise d_revalidate is asking us | |
386 | * to return a fail status. | |
387 | */ | |
388 | if (!status) { | |
389 | if (!d_invalidate(dentry)) { | |
390 | dput(dentry); | |
391 | dentry = NULL; | |
392 | } | |
393 | } else { | |
394 | dput(dentry); | |
395 | dentry = ERR_PTR(status); | |
396 | } | |
397 | } | |
398 | return dentry; | |
399 | } | |
400 | ||
1da177e4 LT |
401 | /* |
402 | * Internal lookup() using the new generic dcache. | |
403 | * SMP-safe | |
404 | */ | |
405 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
406 | { | |
407 | struct dentry * dentry = __d_lookup(parent, name); | |
408 | ||
409 | /* lockess __d_lookup may fail due to concurrent d_move() | |
410 | * in some unrelated directory, so try with d_lookup | |
411 | */ | |
412 | if (!dentry) | |
413 | dentry = d_lookup(parent, name); | |
414 | ||
bcdc5e01 IK |
415 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
416 | dentry = do_revalidate(dentry, nd); | |
417 | ||
1da177e4 LT |
418 | return dentry; |
419 | } | |
420 | ||
421 | /* | |
422 | * Short-cut version of permission(), for calling by | |
423 | * path_walk(), when dcache lock is held. Combines parts | |
424 | * of permission() and generic_permission(), and tests ONLY for | |
425 | * MAY_EXEC permission. | |
426 | * | |
427 | * If appropriate, check DAC only. If not appropriate, or | |
428 | * short-cut DAC fails, then call permission() to do more | |
429 | * complete permission check. | |
430 | */ | |
858119e1 | 431 | static int exec_permission_lite(struct inode *inode, |
1da177e4 LT |
432 | struct nameidata *nd) |
433 | { | |
434 | umode_t mode = inode->i_mode; | |
435 | ||
436 | if (inode->i_op && inode->i_op->permission) | |
437 | return -EAGAIN; | |
438 | ||
439 | if (current->fsuid == inode->i_uid) | |
440 | mode >>= 6; | |
441 | else if (in_group_p(inode->i_gid)) | |
442 | mode >>= 3; | |
443 | ||
444 | if (mode & MAY_EXEC) | |
445 | goto ok; | |
446 | ||
447 | if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE)) | |
448 | goto ok; | |
449 | ||
450 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE)) | |
451 | goto ok; | |
452 | ||
453 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH)) | |
454 | goto ok; | |
455 | ||
456 | return -EACCES; | |
457 | ok: | |
458 | return security_inode_permission(inode, MAY_EXEC, nd); | |
459 | } | |
460 | ||
461 | /* | |
462 | * This is called when everything else fails, and we actually have | |
463 | * to go to the low-level filesystem to find out what we should do.. | |
464 | * | |
465 | * We get the directory semaphore, and after getting that we also | |
466 | * make sure that nobody added the entry to the dcache in the meantime.. | |
467 | * SMP-safe | |
468 | */ | |
469 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
470 | { | |
471 | struct dentry * result; | |
472 | struct inode *dir = parent->d_inode; | |
473 | ||
1b1dcc1b | 474 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
475 | /* |
476 | * First re-do the cached lookup just in case it was created | |
477 | * while we waited for the directory semaphore.. | |
478 | * | |
479 | * FIXME! This could use version numbering or similar to | |
480 | * avoid unnecessary cache lookups. | |
481 | * | |
482 | * The "dcache_lock" is purely to protect the RCU list walker | |
483 | * from concurrent renames at this point (we mustn't get false | |
484 | * negatives from the RCU list walk here, unlike the optimistic | |
485 | * fast walk). | |
486 | * | |
487 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
488 | */ | |
489 | result = d_lookup(parent, name); | |
490 | if (!result) { | |
491 | struct dentry * dentry = d_alloc(parent, name); | |
492 | result = ERR_PTR(-ENOMEM); | |
493 | if (dentry) { | |
494 | result = dir->i_op->lookup(dir, dentry, nd); | |
495 | if (result) | |
496 | dput(dentry); | |
497 | else | |
498 | result = dentry; | |
499 | } | |
1b1dcc1b | 500 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
501 | return result; |
502 | } | |
503 | ||
504 | /* | |
505 | * Uhhuh! Nasty case: the cache was re-populated while | |
506 | * we waited on the semaphore. Need to revalidate. | |
507 | */ | |
1b1dcc1b | 508 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 509 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
510 | result = do_revalidate(result, nd); |
511 | if (!result) | |
1da177e4 | 512 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
513 | } |
514 | return result; | |
515 | } | |
516 | ||
517 | static int __emul_lookup_dentry(const char *, struct nameidata *); | |
518 | ||
519 | /* SMP-safe */ | |
f1662356 | 520 | static __always_inline int |
1da177e4 LT |
521 | walk_init_root(const char *name, struct nameidata *nd) |
522 | { | |
e518ddb7 AM |
523 | struct fs_struct *fs = current->fs; |
524 | ||
525 | read_lock(&fs->lock); | |
526 | if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) { | |
527 | nd->mnt = mntget(fs->altrootmnt); | |
528 | nd->dentry = dget(fs->altroot); | |
529 | read_unlock(&fs->lock); | |
1da177e4 LT |
530 | if (__emul_lookup_dentry(name,nd)) |
531 | return 0; | |
e518ddb7 | 532 | read_lock(&fs->lock); |
1da177e4 | 533 | } |
e518ddb7 AM |
534 | nd->mnt = mntget(fs->rootmnt); |
535 | nd->dentry = dget(fs->root); | |
536 | read_unlock(&fs->lock); | |
1da177e4 LT |
537 | return 1; |
538 | } | |
539 | ||
f1662356 | 540 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
541 | { |
542 | int res = 0; | |
543 | char *name; | |
544 | if (IS_ERR(link)) | |
545 | goto fail; | |
546 | ||
547 | if (*link == '/') { | |
548 | path_release(nd); | |
549 | if (!walk_init_root(link, nd)) | |
550 | /* weird __emul_prefix() stuff did it */ | |
551 | goto out; | |
552 | } | |
553 | res = link_path_walk(link, nd); | |
554 | out: | |
555 | if (nd->depth || res || nd->last_type!=LAST_NORM) | |
556 | return res; | |
557 | /* | |
558 | * If it is an iterative symlinks resolution in open_namei() we | |
559 | * have to copy the last component. And all that crap because of | |
560 | * bloody create() on broken symlinks. Furrfu... | |
561 | */ | |
562 | name = __getname(); | |
563 | if (unlikely(!name)) { | |
564 | path_release(nd); | |
565 | return -ENOMEM; | |
566 | } | |
567 | strcpy(name, nd->last.name); | |
568 | nd->last.name = name; | |
569 | return 0; | |
570 | fail: | |
571 | path_release(nd); | |
572 | return PTR_ERR(link); | |
573 | } | |
574 | ||
051d3812 IK |
575 | static inline void dput_path(struct path *path, struct nameidata *nd) |
576 | { | |
577 | dput(path->dentry); | |
578 | if (path->mnt != nd->mnt) | |
579 | mntput(path->mnt); | |
580 | } | |
581 | ||
582 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
583 | { | |
584 | dput(nd->dentry); | |
585 | if (nd->mnt != path->mnt) | |
586 | mntput(nd->mnt); | |
587 | nd->mnt = path->mnt; | |
588 | nd->dentry = path->dentry; | |
589 | } | |
590 | ||
f1662356 | 591 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
592 | { |
593 | int error; | |
cc314eef | 594 | void *cookie; |
cd4e91d3 | 595 | struct dentry *dentry = path->dentry; |
1da177e4 | 596 | |
d671a1cb | 597 | touch_atime(path->mnt, dentry); |
1da177e4 | 598 | nd_set_link(nd, NULL); |
cd4e91d3 | 599 | |
051d3812 IK |
600 | if (path->mnt != nd->mnt) { |
601 | path_to_nameidata(path, nd); | |
602 | dget(dentry); | |
603 | } | |
604 | mntget(path->mnt); | |
cc314eef LT |
605 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
606 | error = PTR_ERR(cookie); | |
607 | if (!IS_ERR(cookie)) { | |
1da177e4 | 608 | char *s = nd_get_link(nd); |
cc314eef | 609 | error = 0; |
1da177e4 LT |
610 | if (s) |
611 | error = __vfs_follow_link(nd, s); | |
612 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 613 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 614 | } |
cd4e91d3 AV |
615 | dput(dentry); |
616 | mntput(path->mnt); | |
1da177e4 LT |
617 | |
618 | return error; | |
619 | } | |
620 | ||
621 | /* | |
622 | * This limits recursive symlink follows to 8, while | |
623 | * limiting consecutive symlinks to 40. | |
624 | * | |
625 | * Without that kind of total limit, nasty chains of consecutive | |
626 | * symlinks can cause almost arbitrarily long lookups. | |
627 | */ | |
90ebe565 | 628 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
629 | { |
630 | int err = -ELOOP; | |
631 | if (current->link_count >= MAX_NESTED_LINKS) | |
632 | goto loop; | |
633 | if (current->total_link_count >= 40) | |
634 | goto loop; | |
635 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
636 | cond_resched(); | |
90ebe565 | 637 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
638 | if (err) |
639 | goto loop; | |
640 | current->link_count++; | |
641 | current->total_link_count++; | |
642 | nd->depth++; | |
cd4e91d3 | 643 | err = __do_follow_link(path, nd); |
839d9f93 AV |
644 | current->link_count--; |
645 | nd->depth--; | |
1da177e4 LT |
646 | return err; |
647 | loop: | |
09dd17d3 | 648 | dput_path(path, nd); |
839d9f93 | 649 | path_release(nd); |
1da177e4 LT |
650 | return err; |
651 | } | |
652 | ||
653 | int follow_up(struct vfsmount **mnt, struct dentry **dentry) | |
654 | { | |
655 | struct vfsmount *parent; | |
656 | struct dentry *mountpoint; | |
657 | spin_lock(&vfsmount_lock); | |
658 | parent=(*mnt)->mnt_parent; | |
659 | if (parent == *mnt) { | |
660 | spin_unlock(&vfsmount_lock); | |
661 | return 0; | |
662 | } | |
663 | mntget(parent); | |
664 | mountpoint=dget((*mnt)->mnt_mountpoint); | |
665 | spin_unlock(&vfsmount_lock); | |
666 | dput(*dentry); | |
667 | *dentry = mountpoint; | |
668 | mntput(*mnt); | |
669 | *mnt = parent; | |
670 | return 1; | |
671 | } | |
672 | ||
673 | /* no need for dcache_lock, as serialization is taken care in | |
674 | * namespace.c | |
675 | */ | |
463ffb2e AV |
676 | static int __follow_mount(struct path *path) |
677 | { | |
678 | int res = 0; | |
679 | while (d_mountpoint(path->dentry)) { | |
680 | struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry); | |
681 | if (!mounted) | |
682 | break; | |
683 | dput(path->dentry); | |
684 | if (res) | |
685 | mntput(path->mnt); | |
686 | path->mnt = mounted; | |
687 | path->dentry = dget(mounted->mnt_root); | |
688 | res = 1; | |
689 | } | |
690 | return res; | |
691 | } | |
692 | ||
58c465eb | 693 | static void follow_mount(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 | 694 | { |
1da177e4 LT |
695 | while (d_mountpoint(*dentry)) { |
696 | struct vfsmount *mounted = lookup_mnt(*mnt, *dentry); | |
697 | if (!mounted) | |
698 | break; | |
58c465eb | 699 | dput(*dentry); |
1da177e4 LT |
700 | mntput(*mnt); |
701 | *mnt = mounted; | |
1da177e4 | 702 | *dentry = dget(mounted->mnt_root); |
1da177e4 | 703 | } |
1da177e4 LT |
704 | } |
705 | ||
706 | /* no need for dcache_lock, as serialization is taken care in | |
707 | * namespace.c | |
708 | */ | |
e13b210f | 709 | int follow_down(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 LT |
710 | { |
711 | struct vfsmount *mounted; | |
712 | ||
713 | mounted = lookup_mnt(*mnt, *dentry); | |
714 | if (mounted) { | |
e13b210f | 715 | dput(*dentry); |
1da177e4 LT |
716 | mntput(*mnt); |
717 | *mnt = mounted; | |
1da177e4 LT |
718 | *dentry = dget(mounted->mnt_root); |
719 | return 1; | |
720 | } | |
721 | return 0; | |
722 | } | |
723 | ||
f1662356 | 724 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 725 | { |
e518ddb7 AM |
726 | struct fs_struct *fs = current->fs; |
727 | ||
1da177e4 LT |
728 | while(1) { |
729 | struct vfsmount *parent; | |
58c465eb | 730 | struct dentry *old = nd->dentry; |
1da177e4 | 731 | |
e518ddb7 AM |
732 | read_lock(&fs->lock); |
733 | if (nd->dentry == fs->root && | |
734 | nd->mnt == fs->rootmnt) { | |
735 | read_unlock(&fs->lock); | |
1da177e4 LT |
736 | break; |
737 | } | |
e518ddb7 | 738 | read_unlock(&fs->lock); |
1da177e4 | 739 | spin_lock(&dcache_lock); |
58c465eb AV |
740 | if (nd->dentry != nd->mnt->mnt_root) { |
741 | nd->dentry = dget(nd->dentry->d_parent); | |
1da177e4 LT |
742 | spin_unlock(&dcache_lock); |
743 | dput(old); | |
744 | break; | |
745 | } | |
746 | spin_unlock(&dcache_lock); | |
747 | spin_lock(&vfsmount_lock); | |
58c465eb AV |
748 | parent = nd->mnt->mnt_parent; |
749 | if (parent == nd->mnt) { | |
1da177e4 LT |
750 | spin_unlock(&vfsmount_lock); |
751 | break; | |
752 | } | |
753 | mntget(parent); | |
58c465eb | 754 | nd->dentry = dget(nd->mnt->mnt_mountpoint); |
1da177e4 LT |
755 | spin_unlock(&vfsmount_lock); |
756 | dput(old); | |
58c465eb AV |
757 | mntput(nd->mnt); |
758 | nd->mnt = parent; | |
1da177e4 | 759 | } |
58c465eb | 760 | follow_mount(&nd->mnt, &nd->dentry); |
1da177e4 LT |
761 | } |
762 | ||
1da177e4 LT |
763 | /* |
764 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
765 | * small and for now I'd prefer to have fast path as straight as possible. | |
766 | * It _is_ time-critical. | |
767 | */ | |
768 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
769 | struct path *path) | |
770 | { | |
771 | struct vfsmount *mnt = nd->mnt; | |
772 | struct dentry *dentry = __d_lookup(nd->dentry, name); | |
773 | ||
774 | if (!dentry) | |
775 | goto need_lookup; | |
776 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
777 | goto need_revalidate; | |
778 | done: | |
779 | path->mnt = mnt; | |
780 | path->dentry = dentry; | |
634ee701 | 781 | __follow_mount(path); |
1da177e4 LT |
782 | return 0; |
783 | ||
784 | need_lookup: | |
785 | dentry = real_lookup(nd->dentry, name, nd); | |
786 | if (IS_ERR(dentry)) | |
787 | goto fail; | |
788 | goto done; | |
789 | ||
790 | need_revalidate: | |
bcdc5e01 IK |
791 | dentry = do_revalidate(dentry, nd); |
792 | if (!dentry) | |
793 | goto need_lookup; | |
794 | if (IS_ERR(dentry)) | |
795 | goto fail; | |
796 | goto done; | |
1da177e4 LT |
797 | |
798 | fail: | |
799 | return PTR_ERR(dentry); | |
800 | } | |
801 | ||
802 | /* | |
803 | * Name resolution. | |
ea3834d9 PM |
804 | * This is the basic name resolution function, turning a pathname into |
805 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 806 | * |
ea3834d9 PM |
807 | * Returns 0 and nd will have valid dentry and mnt on success. |
808 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 LT |
809 | */ |
810 | static fastcall int __link_path_walk(const char * name, struct nameidata *nd) | |
811 | { | |
812 | struct path next; | |
813 | struct inode *inode; | |
814 | int err; | |
815 | unsigned int lookup_flags = nd->flags; | |
816 | ||
817 | while (*name=='/') | |
818 | name++; | |
819 | if (!*name) | |
820 | goto return_reval; | |
821 | ||
822 | inode = nd->dentry->d_inode; | |
823 | if (nd->depth) | |
f55eab82 | 824 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
825 | |
826 | /* At this point we know we have a real path component. */ | |
827 | for(;;) { | |
828 | unsigned long hash; | |
829 | struct qstr this; | |
830 | unsigned int c; | |
831 | ||
cdce5d6b | 832 | nd->flags |= LOOKUP_CONTINUE; |
1da177e4 | 833 | err = exec_permission_lite(inode, nd); |
e4543edd CH |
834 | if (err == -EAGAIN) |
835 | err = vfs_permission(nd, MAY_EXEC); | |
1da177e4 LT |
836 | if (err) |
837 | break; | |
838 | ||
839 | this.name = name; | |
840 | c = *(const unsigned char *)name; | |
841 | ||
842 | hash = init_name_hash(); | |
843 | do { | |
844 | name++; | |
845 | hash = partial_name_hash(c, hash); | |
846 | c = *(const unsigned char *)name; | |
847 | } while (c && (c != '/')); | |
848 | this.len = name - (const char *) this.name; | |
849 | this.hash = end_name_hash(hash); | |
850 | ||
851 | /* remove trailing slashes? */ | |
852 | if (!c) | |
853 | goto last_component; | |
854 | while (*++name == '/'); | |
855 | if (!*name) | |
856 | goto last_with_slashes; | |
857 | ||
858 | /* | |
859 | * "." and ".." are special - ".." especially so because it has | |
860 | * to be able to know about the current root directory and | |
861 | * parent relationships. | |
862 | */ | |
863 | if (this.name[0] == '.') switch (this.len) { | |
864 | default: | |
865 | break; | |
866 | case 2: | |
867 | if (this.name[1] != '.') | |
868 | break; | |
58c465eb | 869 | follow_dotdot(nd); |
1da177e4 LT |
870 | inode = nd->dentry->d_inode; |
871 | /* fallthrough */ | |
872 | case 1: | |
873 | continue; | |
874 | } | |
875 | /* | |
876 | * See if the low-level filesystem might want | |
877 | * to use its own hash.. | |
878 | */ | |
879 | if (nd->dentry->d_op && nd->dentry->d_op->d_hash) { | |
880 | err = nd->dentry->d_op->d_hash(nd->dentry, &this); | |
881 | if (err < 0) | |
882 | break; | |
883 | } | |
1da177e4 LT |
884 | /* This does the actual lookups.. */ |
885 | err = do_lookup(nd, &this, &next); | |
886 | if (err) | |
887 | break; | |
1da177e4 LT |
888 | |
889 | err = -ENOENT; | |
890 | inode = next.dentry->d_inode; | |
891 | if (!inode) | |
892 | goto out_dput; | |
893 | err = -ENOTDIR; | |
894 | if (!inode->i_op) | |
895 | goto out_dput; | |
896 | ||
897 | if (inode->i_op->follow_link) { | |
90ebe565 | 898 | err = do_follow_link(&next, nd); |
1da177e4 LT |
899 | if (err) |
900 | goto return_err; | |
901 | err = -ENOENT; | |
902 | inode = nd->dentry->d_inode; | |
903 | if (!inode) | |
904 | break; | |
905 | err = -ENOTDIR; | |
906 | if (!inode->i_op) | |
907 | break; | |
09dd17d3 MS |
908 | } else |
909 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
910 | err = -ENOTDIR; |
911 | if (!inode->i_op->lookup) | |
912 | break; | |
913 | continue; | |
914 | /* here ends the main loop */ | |
915 | ||
916 | last_with_slashes: | |
917 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
918 | last_component: | |
f55eab82 TM |
919 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
920 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
921 | if (lookup_flags & LOOKUP_PARENT) |
922 | goto lookup_parent; | |
923 | if (this.name[0] == '.') switch (this.len) { | |
924 | default: | |
925 | break; | |
926 | case 2: | |
927 | if (this.name[1] != '.') | |
928 | break; | |
58c465eb | 929 | follow_dotdot(nd); |
1da177e4 LT |
930 | inode = nd->dentry->d_inode; |
931 | /* fallthrough */ | |
932 | case 1: | |
933 | goto return_reval; | |
934 | } | |
935 | if (nd->dentry->d_op && nd->dentry->d_op->d_hash) { | |
936 | err = nd->dentry->d_op->d_hash(nd->dentry, &this); | |
937 | if (err < 0) | |
938 | break; | |
939 | } | |
940 | err = do_lookup(nd, &this, &next); | |
941 | if (err) | |
942 | break; | |
1da177e4 LT |
943 | inode = next.dentry->d_inode; |
944 | if ((lookup_flags & LOOKUP_FOLLOW) | |
945 | && inode && inode->i_op && inode->i_op->follow_link) { | |
90ebe565 | 946 | err = do_follow_link(&next, nd); |
1da177e4 LT |
947 | if (err) |
948 | goto return_err; | |
949 | inode = nd->dentry->d_inode; | |
09dd17d3 MS |
950 | } else |
951 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
952 | err = -ENOENT; |
953 | if (!inode) | |
954 | break; | |
955 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
956 | err = -ENOTDIR; | |
957 | if (!inode->i_op || !inode->i_op->lookup) | |
958 | break; | |
959 | } | |
960 | goto return_base; | |
961 | lookup_parent: | |
962 | nd->last = this; | |
963 | nd->last_type = LAST_NORM; | |
964 | if (this.name[0] != '.') | |
965 | goto return_base; | |
966 | if (this.len == 1) | |
967 | nd->last_type = LAST_DOT; | |
968 | else if (this.len == 2 && this.name[1] == '.') | |
969 | nd->last_type = LAST_DOTDOT; | |
970 | else | |
971 | goto return_base; | |
972 | return_reval: | |
973 | /* | |
974 | * We bypassed the ordinary revalidation routines. | |
975 | * We may need to check the cached dentry for staleness. | |
976 | */ | |
977 | if (nd->dentry && nd->dentry->d_sb && | |
978 | (nd->dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
979 | err = -ESTALE; | |
980 | /* Note: we do not d_invalidate() */ | |
981 | if (!nd->dentry->d_op->d_revalidate(nd->dentry, nd)) | |
982 | break; | |
983 | } | |
984 | return_base: | |
985 | return 0; | |
986 | out_dput: | |
09dd17d3 | 987 | dput_path(&next, nd); |
1da177e4 LT |
988 | break; |
989 | } | |
990 | path_release(nd); | |
991 | return_err: | |
992 | return err; | |
993 | } | |
994 | ||
995 | /* | |
996 | * Wrapper to retry pathname resolution whenever the underlying | |
997 | * file system returns an ESTALE. | |
998 | * | |
999 | * Retry the whole path once, forcing real lookup requests | |
1000 | * instead of relying on the dcache. | |
1001 | */ | |
1002 | int fastcall link_path_walk(const char *name, struct nameidata *nd) | |
1003 | { | |
1004 | struct nameidata save = *nd; | |
1005 | int result; | |
1006 | ||
1007 | /* make sure the stuff we saved doesn't go away */ | |
1008 | dget(save.dentry); | |
1009 | mntget(save.mnt); | |
1010 | ||
1011 | result = __link_path_walk(name, nd); | |
1012 | if (result == -ESTALE) { | |
1013 | *nd = save; | |
1014 | dget(nd->dentry); | |
1015 | mntget(nd->mnt); | |
1016 | nd->flags |= LOOKUP_REVAL; | |
1017 | result = __link_path_walk(name, nd); | |
1018 | } | |
1019 | ||
1020 | dput(save.dentry); | |
1021 | mntput(save.mnt); | |
1022 | ||
1023 | return result; | |
1024 | } | |
1025 | ||
1026 | int fastcall path_walk(const char * name, struct nameidata *nd) | |
1027 | { | |
1028 | current->total_link_count = 0; | |
1029 | return link_path_walk(name, nd); | |
1030 | } | |
1031 | ||
ea3834d9 PM |
1032 | /* |
1033 | * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if | |
1034 | * everything is done. Returns 0 and drops input nd, if lookup failed; | |
1035 | */ | |
1da177e4 LT |
1036 | static int __emul_lookup_dentry(const char *name, struct nameidata *nd) |
1037 | { | |
1038 | if (path_walk(name, nd)) | |
1039 | return 0; /* something went wrong... */ | |
1040 | ||
1041 | if (!nd->dentry->d_inode || S_ISDIR(nd->dentry->d_inode->i_mode)) { | |
1042 | struct dentry *old_dentry = nd->dentry; | |
1043 | struct vfsmount *old_mnt = nd->mnt; | |
1044 | struct qstr last = nd->last; | |
1045 | int last_type = nd->last_type; | |
e518ddb7 AM |
1046 | struct fs_struct *fs = current->fs; |
1047 | ||
1da177e4 | 1048 | /* |
e518ddb7 AM |
1049 | * NAME was not found in alternate root or it's a directory. |
1050 | * Try to find it in the normal root: | |
1da177e4 LT |
1051 | */ |
1052 | nd->last_type = LAST_ROOT; | |
e518ddb7 AM |
1053 | read_lock(&fs->lock); |
1054 | nd->mnt = mntget(fs->rootmnt); | |
1055 | nd->dentry = dget(fs->root); | |
1056 | read_unlock(&fs->lock); | |
1da177e4 LT |
1057 | if (path_walk(name, nd) == 0) { |
1058 | if (nd->dentry->d_inode) { | |
1059 | dput(old_dentry); | |
1060 | mntput(old_mnt); | |
1061 | return 1; | |
1062 | } | |
1063 | path_release(nd); | |
1064 | } | |
1065 | nd->dentry = old_dentry; | |
1066 | nd->mnt = old_mnt; | |
1067 | nd->last = last; | |
1068 | nd->last_type = last_type; | |
1069 | } | |
1070 | return 1; | |
1071 | } | |
1072 | ||
1073 | void set_fs_altroot(void) | |
1074 | { | |
1075 | char *emul = __emul_prefix(); | |
1076 | struct nameidata nd; | |
1077 | struct vfsmount *mnt = NULL, *oldmnt; | |
1078 | struct dentry *dentry = NULL, *olddentry; | |
1079 | int err; | |
e518ddb7 | 1080 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1081 | |
1082 | if (!emul) | |
1083 | goto set_it; | |
1084 | err = path_lookup(emul, LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_NOALT, &nd); | |
1085 | if (!err) { | |
1086 | mnt = nd.mnt; | |
1087 | dentry = nd.dentry; | |
1088 | } | |
1089 | set_it: | |
e518ddb7 AM |
1090 | write_lock(&fs->lock); |
1091 | oldmnt = fs->altrootmnt; | |
1092 | olddentry = fs->altroot; | |
1093 | fs->altrootmnt = mnt; | |
1094 | fs->altroot = dentry; | |
1095 | write_unlock(&fs->lock); | |
1da177e4 LT |
1096 | if (olddentry) { |
1097 | dput(olddentry); | |
1098 | mntput(oldmnt); | |
1099 | } | |
1100 | } | |
1101 | ||
ea3834d9 | 1102 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
5590ff0d UD |
1103 | static int fastcall do_path_lookup(int dfd, const char *name, |
1104 | unsigned int flags, struct nameidata *nd) | |
1da177e4 | 1105 | { |
ea3834d9 | 1106 | int retval = 0; |
170aa3d0 UD |
1107 | int fput_needed; |
1108 | struct file *file; | |
e518ddb7 | 1109 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1110 | |
1111 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1112 | nd->flags = flags; | |
1113 | nd->depth = 0; | |
1114 | ||
1da177e4 | 1115 | if (*name=='/') { |
e518ddb7 AM |
1116 | read_lock(&fs->lock); |
1117 | if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) { | |
1118 | nd->mnt = mntget(fs->altrootmnt); | |
1119 | nd->dentry = dget(fs->altroot); | |
1120 | read_unlock(&fs->lock); | |
1da177e4 | 1121 | if (__emul_lookup_dentry(name,nd)) |
ea3834d9 | 1122 | goto out; /* found in altroot */ |
e518ddb7 | 1123 | read_lock(&fs->lock); |
1da177e4 | 1124 | } |
e518ddb7 AM |
1125 | nd->mnt = mntget(fs->rootmnt); |
1126 | nd->dentry = dget(fs->root); | |
1127 | read_unlock(&fs->lock); | |
5590ff0d | 1128 | } else if (dfd == AT_FDCWD) { |
e518ddb7 AM |
1129 | read_lock(&fs->lock); |
1130 | nd->mnt = mntget(fs->pwdmnt); | |
1131 | nd->dentry = dget(fs->pwd); | |
1132 | read_unlock(&fs->lock); | |
5590ff0d | 1133 | } else { |
5590ff0d UD |
1134 | struct dentry *dentry; |
1135 | ||
1136 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1137 | retval = -EBADF; |
1138 | if (!file) | |
6d09bb62 | 1139 | goto out_fail; |
5590ff0d | 1140 | |
0f7fc9e4 | 1141 | dentry = file->f_path.dentry; |
5590ff0d | 1142 | |
170aa3d0 UD |
1143 | retval = -ENOTDIR; |
1144 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1145 | goto fput_fail; |
5590ff0d UD |
1146 | |
1147 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1148 | if (retval) |
6d09bb62 | 1149 | goto fput_fail; |
5590ff0d | 1150 | |
0f7fc9e4 | 1151 | nd->mnt = mntget(file->f_path.mnt); |
5590ff0d UD |
1152 | nd->dentry = dget(dentry); |
1153 | ||
1154 | fput_light(file, fput_needed); | |
1da177e4 | 1155 | } |
1da177e4 LT |
1156 | current->total_link_count = 0; |
1157 | retval = link_path_walk(name, nd); | |
ea3834d9 | 1158 | out: |
3bc8414b | 1159 | if (likely(retval == 0)) { |
5ac3a9c2 | 1160 | if (unlikely(!audit_dummy_context() && nd && nd->dentry && |
3bc8414b | 1161 | nd->dentry->d_inode)) |
9c937dcc | 1162 | audit_inode(name, nd->dentry->d_inode); |
3bc8414b | 1163 | } |
6d09bb62 | 1164 | out_fail: |
170aa3d0 UD |
1165 | return retval; |
1166 | ||
6d09bb62 | 1167 | fput_fail: |
170aa3d0 | 1168 | fput_light(file, fput_needed); |
6d09bb62 | 1169 | goto out_fail; |
1da177e4 LT |
1170 | } |
1171 | ||
5590ff0d UD |
1172 | int fastcall path_lookup(const char *name, unsigned int flags, |
1173 | struct nameidata *nd) | |
1174 | { | |
1175 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1176 | } | |
1177 | ||
1178 | static int __path_lookup_intent_open(int dfd, const char *name, | |
1179 | unsigned int lookup_flags, struct nameidata *nd, | |
1180 | int open_flags, int create_mode) | |
834f2a4a TM |
1181 | { |
1182 | struct file *filp = get_empty_filp(); | |
1183 | int err; | |
1184 | ||
1185 | if (filp == NULL) | |
1186 | return -ENFILE; | |
1187 | nd->intent.open.file = filp; | |
1188 | nd->intent.open.flags = open_flags; | |
1189 | nd->intent.open.create_mode = create_mode; | |
5590ff0d | 1190 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1191 | if (IS_ERR(nd->intent.open.file)) { |
1192 | if (err == 0) { | |
1193 | err = PTR_ERR(nd->intent.open.file); | |
1194 | path_release(nd); | |
1195 | } | |
1196 | } else if (err != 0) | |
1197 | release_open_intent(nd); | |
1198 | return err; | |
1199 | } | |
1200 | ||
1201 | /** | |
1202 | * path_lookup_open - lookup a file path with open intent | |
7045f37b | 1203 | * @dfd: the directory to use as base, or AT_FDCWD |
834f2a4a TM |
1204 | * @name: pointer to file name |
1205 | * @lookup_flags: lookup intent flags | |
1206 | * @nd: pointer to nameidata | |
1207 | * @open_flags: open intent flags | |
1208 | */ | |
5590ff0d | 1209 | int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags, |
834f2a4a TM |
1210 | struct nameidata *nd, int open_flags) |
1211 | { | |
5590ff0d | 1212 | return __path_lookup_intent_open(dfd, name, lookup_flags, nd, |
834f2a4a TM |
1213 | open_flags, 0); |
1214 | } | |
1215 | ||
1216 | /** | |
1217 | * path_lookup_create - lookup a file path with open + create intent | |
7045f37b | 1218 | * @dfd: the directory to use as base, or AT_FDCWD |
834f2a4a TM |
1219 | * @name: pointer to file name |
1220 | * @lookup_flags: lookup intent flags | |
1221 | * @nd: pointer to nameidata | |
1222 | * @open_flags: open intent flags | |
1223 | * @create_mode: create intent flags | |
1224 | */ | |
5590ff0d UD |
1225 | static int path_lookup_create(int dfd, const char *name, |
1226 | unsigned int lookup_flags, struct nameidata *nd, | |
1227 | int open_flags, int create_mode) | |
834f2a4a | 1228 | { |
5590ff0d UD |
1229 | return __path_lookup_intent_open(dfd, name, lookup_flags|LOOKUP_CREATE, |
1230 | nd, open_flags, create_mode); | |
834f2a4a TM |
1231 | } |
1232 | ||
1233 | int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags, | |
1234 | struct nameidata *nd, int open_flags) | |
1235 | { | |
1236 | char *tmp = getname(name); | |
1237 | int err = PTR_ERR(tmp); | |
1238 | ||
1239 | if (!IS_ERR(tmp)) { | |
5590ff0d | 1240 | err = __path_lookup_intent_open(AT_FDCWD, tmp, lookup_flags, nd, open_flags, 0); |
834f2a4a TM |
1241 | putname(tmp); |
1242 | } | |
1243 | return err; | |
1244 | } | |
1245 | ||
057f6c01 | 1246 | static inline struct dentry *__lookup_hash_kern(struct qstr *name, struct dentry *base, struct nameidata *nd) |
1da177e4 | 1247 | { |
057f6c01 | 1248 | struct dentry *dentry; |
1da177e4 LT |
1249 | struct inode *inode; |
1250 | int err; | |
1251 | ||
1252 | inode = base->d_inode; | |
1da177e4 LT |
1253 | |
1254 | /* | |
1255 | * See if the low-level filesystem might want | |
1256 | * to use its own hash.. | |
1257 | */ | |
1258 | if (base->d_op && base->d_op->d_hash) { | |
1259 | err = base->d_op->d_hash(base, name); | |
1260 | dentry = ERR_PTR(err); | |
1261 | if (err < 0) | |
1262 | goto out; | |
1263 | } | |
1264 | ||
1265 | dentry = cached_lookup(base, name, nd); | |
1266 | if (!dentry) { | |
1267 | struct dentry *new = d_alloc(base, name); | |
1268 | dentry = ERR_PTR(-ENOMEM); | |
1269 | if (!new) | |
1270 | goto out; | |
1271 | dentry = inode->i_op->lookup(inode, new, nd); | |
1272 | if (!dentry) | |
1273 | dentry = new; | |
1274 | else | |
1275 | dput(new); | |
1276 | } | |
1277 | out: | |
1278 | return dentry; | |
1279 | } | |
1280 | ||
057f6c01 JM |
1281 | /* |
1282 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1283 | * needs parent already locked. Doesn't follow mounts. | |
1284 | * SMP-safe. | |
1285 | */ | |
1286 | static inline struct dentry * __lookup_hash(struct qstr *name, struct dentry *base, struct nameidata *nd) | |
1287 | { | |
1288 | struct dentry *dentry; | |
1289 | struct inode *inode; | |
1290 | int err; | |
1291 | ||
1292 | inode = base->d_inode; | |
1293 | ||
1294 | err = permission(inode, MAY_EXEC, nd); | |
1295 | dentry = ERR_PTR(err); | |
1296 | if (err) | |
1297 | goto out; | |
1298 | ||
1299 | dentry = __lookup_hash_kern(name, base, nd); | |
1300 | out: | |
1301 | return dentry; | |
1302 | } | |
1303 | ||
a244e169 | 1304 | static struct dentry *lookup_hash(struct nameidata *nd) |
1da177e4 | 1305 | { |
49705b77 | 1306 | return __lookup_hash(&nd->last, nd->dentry, nd); |
1da177e4 LT |
1307 | } |
1308 | ||
1309 | /* SMP-safe */ | |
057f6c01 | 1310 | static inline int __lookup_one_len(const char *name, struct qstr *this, struct dentry *base, int len) |
1da177e4 LT |
1311 | { |
1312 | unsigned long hash; | |
1da177e4 LT |
1313 | unsigned int c; |
1314 | ||
057f6c01 JM |
1315 | this->name = name; |
1316 | this->len = len; | |
1da177e4 | 1317 | if (!len) |
057f6c01 | 1318 | return -EACCES; |
1da177e4 LT |
1319 | |
1320 | hash = init_name_hash(); | |
1321 | while (len--) { | |
1322 | c = *(const unsigned char *)name++; | |
1323 | if (c == '/' || c == '\0') | |
057f6c01 | 1324 | return -EACCES; |
1da177e4 LT |
1325 | hash = partial_name_hash(c, hash); |
1326 | } | |
057f6c01 JM |
1327 | this->hash = end_name_hash(hash); |
1328 | return 0; | |
1329 | } | |
1da177e4 | 1330 | |
057f6c01 JM |
1331 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1332 | { | |
1333 | int err; | |
1334 | struct qstr this; | |
1335 | ||
1336 | err = __lookup_one_len(name, &this, base, len); | |
1337 | if (err) | |
1338 | return ERR_PTR(err); | |
49705b77 | 1339 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1340 | } |
1341 | ||
1342 | struct dentry *lookup_one_len_kern(const char *name, struct dentry *base, int len) | |
1343 | { | |
1344 | int err; | |
1345 | struct qstr this; | |
1346 | ||
1347 | err = __lookup_one_len(name, &this, base, len); | |
1348 | if (err) | |
1349 | return ERR_PTR(err); | |
1350 | return __lookup_hash_kern(&this, base, NULL); | |
1da177e4 LT |
1351 | } |
1352 | ||
1353 | /* | |
1354 | * namei() | |
1355 | * | |
1356 | * is used by most simple commands to get the inode of a specified name. | |
1357 | * Open, link etc use their own routines, but this is enough for things | |
1358 | * like 'chmod' etc. | |
1359 | * | |
1360 | * namei exists in two versions: namei/lnamei. The only difference is | |
1361 | * that namei follows links, while lnamei does not. | |
1362 | * SMP-safe | |
1363 | */ | |
5590ff0d UD |
1364 | int fastcall __user_walk_fd(int dfd, const char __user *name, unsigned flags, |
1365 | struct nameidata *nd) | |
1da177e4 LT |
1366 | { |
1367 | char *tmp = getname(name); | |
1368 | int err = PTR_ERR(tmp); | |
1369 | ||
1370 | if (!IS_ERR(tmp)) { | |
5590ff0d | 1371 | err = do_path_lookup(dfd, tmp, flags, nd); |
1da177e4 LT |
1372 | putname(tmp); |
1373 | } | |
1374 | return err; | |
1375 | } | |
1376 | ||
5590ff0d UD |
1377 | int fastcall __user_walk(const char __user *name, unsigned flags, struct nameidata *nd) |
1378 | { | |
1379 | return __user_walk_fd(AT_FDCWD, name, flags, nd); | |
1380 | } | |
1381 | ||
1da177e4 LT |
1382 | /* |
1383 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1384 | * minimal. | |
1385 | */ | |
1386 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1387 | { | |
1388 | if (!(dir->i_mode & S_ISVTX)) | |
1389 | return 0; | |
1390 | if (inode->i_uid == current->fsuid) | |
1391 | return 0; | |
1392 | if (dir->i_uid == current->fsuid) | |
1393 | return 0; | |
1394 | return !capable(CAP_FOWNER); | |
1395 | } | |
1396 | ||
1397 | /* | |
1398 | * Check whether we can remove a link victim from directory dir, check | |
1399 | * whether the type of victim is right. | |
1400 | * 1. We can't do it if dir is read-only (done in permission()) | |
1401 | * 2. We should have write and exec permissions on dir | |
1402 | * 3. We can't remove anything from append-only dir | |
1403 | * 4. We can't do anything with immutable dir (done in permission()) | |
1404 | * 5. If the sticky bit on dir is set we should either | |
1405 | * a. be owner of dir, or | |
1406 | * b. be owner of victim, or | |
1407 | * c. have CAP_FOWNER capability | |
1408 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1409 | * links pointing to it. | |
1410 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1411 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1412 | * 9. We can't remove a root or mountpoint. | |
1413 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1414 | * nfs_async_unlink(). | |
1415 | */ | |
858119e1 | 1416 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1417 | { |
1418 | int error; | |
1419 | ||
1420 | if (!victim->d_inode) | |
1421 | return -ENOENT; | |
1422 | ||
1423 | BUG_ON(victim->d_parent->d_inode != dir); | |
73d3ec5a | 1424 | audit_inode_child(victim->d_name.name, victim->d_inode, dir); |
1da177e4 LT |
1425 | |
1426 | error = permission(dir,MAY_WRITE | MAY_EXEC, NULL); | |
1427 | if (error) | |
1428 | return error; | |
1429 | if (IS_APPEND(dir)) | |
1430 | return -EPERM; | |
1431 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
1432 | IS_IMMUTABLE(victim->d_inode)) | |
1433 | return -EPERM; | |
1434 | if (isdir) { | |
1435 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1436 | return -ENOTDIR; | |
1437 | if (IS_ROOT(victim)) | |
1438 | return -EBUSY; | |
1439 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1440 | return -EISDIR; | |
1441 | if (IS_DEADDIR(dir)) | |
1442 | return -ENOENT; | |
1443 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1444 | return -EBUSY; | |
1445 | return 0; | |
1446 | } | |
1447 | ||
1448 | /* Check whether we can create an object with dentry child in directory | |
1449 | * dir. | |
1450 | * 1. We can't do it if child already exists (open has special treatment for | |
1451 | * this case, but since we are inlined it's OK) | |
1452 | * 2. We can't do it if dir is read-only (done in permission()) | |
1453 | * 3. We should have write and exec permissions on dir | |
1454 | * 4. We can't do it if dir is immutable (done in permission()) | |
1455 | */ | |
1456 | static inline int may_create(struct inode *dir, struct dentry *child, | |
1457 | struct nameidata *nd) | |
1458 | { | |
1459 | if (child->d_inode) | |
1460 | return -EEXIST; | |
1461 | if (IS_DEADDIR(dir)) | |
1462 | return -ENOENT; | |
1463 | return permission(dir,MAY_WRITE | MAY_EXEC, nd); | |
1464 | } | |
1465 | ||
1466 | /* | |
1da177e4 LT |
1467 | * O_DIRECTORY translates into forcing a directory lookup. |
1468 | */ | |
1469 | static inline int lookup_flags(unsigned int f) | |
1470 | { | |
1471 | unsigned long retval = LOOKUP_FOLLOW; | |
1472 | ||
1473 | if (f & O_NOFOLLOW) | |
1474 | retval &= ~LOOKUP_FOLLOW; | |
1475 | ||
1da177e4 LT |
1476 | if (f & O_DIRECTORY) |
1477 | retval |= LOOKUP_DIRECTORY; | |
1478 | ||
1479 | return retval; | |
1480 | } | |
1481 | ||
1482 | /* | |
1483 | * p1 and p2 should be directories on the same fs. | |
1484 | */ | |
1485 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1486 | { | |
1487 | struct dentry *p; | |
1488 | ||
1489 | if (p1 == p2) { | |
f2eace23 | 1490 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1491 | return NULL; |
1492 | } | |
1493 | ||
a11f3a05 | 1494 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1495 | |
1496 | for (p = p1; p->d_parent != p; p = p->d_parent) { | |
1497 | if (p->d_parent == p2) { | |
f2eace23 IM |
1498 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); |
1499 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1500 | return p; |
1501 | } | |
1502 | } | |
1503 | ||
1504 | for (p = p2; p->d_parent != p; p = p->d_parent) { | |
1505 | if (p->d_parent == p1) { | |
f2eace23 IM |
1506 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1507 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1508 | return p; |
1509 | } | |
1510 | } | |
1511 | ||
f2eace23 IM |
1512 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1513 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1514 | return NULL; |
1515 | } | |
1516 | ||
1517 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1518 | { | |
1b1dcc1b | 1519 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1520 | if (p1 != p2) { |
1b1dcc1b | 1521 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1522 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1523 | } |
1524 | } | |
1525 | ||
1526 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1527 | struct nameidata *nd) | |
1528 | { | |
1529 | int error = may_create(dir, dentry, nd); | |
1530 | ||
1531 | if (error) | |
1532 | return error; | |
1533 | ||
1534 | if (!dir->i_op || !dir->i_op->create) | |
1535 | return -EACCES; /* shouldn't it be ENOSYS? */ | |
1536 | mode &= S_IALLUGO; | |
1537 | mode |= S_IFREG; | |
1538 | error = security_inode_create(dir, dentry, mode); | |
1539 | if (error) | |
1540 | return error; | |
1541 | DQUOT_INIT(dir); | |
1542 | error = dir->i_op->create(dir, dentry, mode, nd); | |
a74574aa | 1543 | if (!error) |
f38aa942 | 1544 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1545 | return error; |
1546 | } | |
1547 | ||
1548 | int may_open(struct nameidata *nd, int acc_mode, int flag) | |
1549 | { | |
1550 | struct dentry *dentry = nd->dentry; | |
1551 | struct inode *inode = dentry->d_inode; | |
1552 | int error; | |
1553 | ||
1554 | if (!inode) | |
1555 | return -ENOENT; | |
1556 | ||
1557 | if (S_ISLNK(inode->i_mode)) | |
1558 | return -ELOOP; | |
1559 | ||
1560 | if (S_ISDIR(inode->i_mode) && (flag & FMODE_WRITE)) | |
1561 | return -EISDIR; | |
1562 | ||
e4543edd | 1563 | error = vfs_permission(nd, acc_mode); |
1da177e4 LT |
1564 | if (error) |
1565 | return error; | |
1566 | ||
1567 | /* | |
1568 | * FIFO's, sockets and device files are special: they don't | |
1569 | * actually live on the filesystem itself, and as such you | |
1570 | * can write to them even if the filesystem is read-only. | |
1571 | */ | |
1572 | if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { | |
1573 | flag &= ~O_TRUNC; | |
1574 | } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) { | |
1575 | if (nd->mnt->mnt_flags & MNT_NODEV) | |
1576 | return -EACCES; | |
1577 | ||
1578 | flag &= ~O_TRUNC; | |
1579 | } else if (IS_RDONLY(inode) && (flag & FMODE_WRITE)) | |
1580 | return -EROFS; | |
1581 | /* | |
1582 | * An append-only file must be opened in append mode for writing. | |
1583 | */ | |
1584 | if (IS_APPEND(inode)) { | |
1585 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
1586 | return -EPERM; | |
1587 | if (flag & O_TRUNC) | |
1588 | return -EPERM; | |
1589 | } | |
1590 | ||
1591 | /* O_NOATIME can only be set by the owner or superuser */ | |
1592 | if (flag & O_NOATIME) | |
1593 | if (current->fsuid != inode->i_uid && !capable(CAP_FOWNER)) | |
1594 | return -EPERM; | |
1595 | ||
1596 | /* | |
1597 | * Ensure there are no outstanding leases on the file. | |
1598 | */ | |
1599 | error = break_lease(inode, flag); | |
1600 | if (error) | |
1601 | return error; | |
1602 | ||
1603 | if (flag & O_TRUNC) { | |
1604 | error = get_write_access(inode); | |
1605 | if (error) | |
1606 | return error; | |
1607 | ||
1608 | /* | |
1609 | * Refuse to truncate files with mandatory locks held on them. | |
1610 | */ | |
1611 | error = locks_verify_locked(inode); | |
1612 | if (!error) { | |
1613 | DQUOT_INIT(inode); | |
1614 | ||
4a30131e | 1615 | error = do_truncate(dentry, 0, ATTR_MTIME|ATTR_CTIME, NULL); |
1da177e4 LT |
1616 | } |
1617 | put_write_access(inode); | |
1618 | if (error) | |
1619 | return error; | |
1620 | } else | |
1621 | if (flag & FMODE_WRITE) | |
1622 | DQUOT_INIT(inode); | |
1623 | ||
1624 | return 0; | |
1625 | } | |
1626 | ||
aab520e2 DH |
1627 | static int open_namei_create(struct nameidata *nd, struct path *path, |
1628 | int flag, int mode) | |
1629 | { | |
1630 | int error; | |
1631 | struct dentry *dir = nd->dentry; | |
1632 | ||
1633 | if (!IS_POSIXACL(dir->d_inode)) | |
1634 | mode &= ~current->fs->umask; | |
1635 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); | |
1636 | mutex_unlock(&dir->d_inode->i_mutex); | |
1637 | dput(nd->dentry); | |
1638 | nd->dentry = path->dentry; | |
1639 | if (error) | |
1640 | return error; | |
1641 | /* Don't check for write permission, don't truncate */ | |
1642 | return may_open(nd, 0, flag & ~O_TRUNC); | |
1643 | } | |
1644 | ||
1da177e4 LT |
1645 | /* |
1646 | * open_namei() | |
1647 | * | |
1648 | * namei for open - this is in fact almost the whole open-routine. | |
1649 | * | |
1650 | * Note that the low bits of "flag" aren't the same as in the open | |
1651 | * system call - they are 00 - no permissions needed | |
1652 | * 01 - read permission needed | |
1653 | * 10 - write permission needed | |
1654 | * 11 - read/write permissions needed | |
1655 | * which is a lot more logical, and also allows the "no perm" needed | |
1656 | * for symlinks (where the permissions are checked later). | |
1657 | * SMP-safe | |
1658 | */ | |
5590ff0d UD |
1659 | int open_namei(int dfd, const char *pathname, int flag, |
1660 | int mode, struct nameidata *nd) | |
1da177e4 | 1661 | { |
834f2a4a | 1662 | int acc_mode, error; |
4e7506e4 | 1663 | struct path path; |
1da177e4 LT |
1664 | struct dentry *dir; |
1665 | int count = 0; | |
1666 | ||
1667 | acc_mode = ACC_MODE(flag); | |
1668 | ||
834f2a4a TM |
1669 | /* O_TRUNC implies we need access checks for write permissions */ |
1670 | if (flag & O_TRUNC) | |
1671 | acc_mode |= MAY_WRITE; | |
1672 | ||
1da177e4 LT |
1673 | /* Allow the LSM permission hook to distinguish append |
1674 | access from general write access. */ | |
1675 | if (flag & O_APPEND) | |
1676 | acc_mode |= MAY_APPEND; | |
1677 | ||
1da177e4 LT |
1678 | /* |
1679 | * The simplest case - just a plain lookup. | |
1680 | */ | |
1681 | if (!(flag & O_CREAT)) { | |
5590ff0d UD |
1682 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
1683 | nd, flag); | |
1da177e4 LT |
1684 | if (error) |
1685 | return error; | |
1686 | goto ok; | |
1687 | } | |
1688 | ||
1689 | /* | |
1690 | * Create - we need to know the parent. | |
1691 | */ | |
5590ff0d | 1692 | error = path_lookup_create(dfd,pathname,LOOKUP_PARENT,nd,flag,mode); |
1da177e4 LT |
1693 | if (error) |
1694 | return error; | |
1695 | ||
1696 | /* | |
1697 | * We have the parent and last component. First of all, check | |
1698 | * that we are not asked to creat(2) an obvious directory - that | |
1699 | * will not do. | |
1700 | */ | |
1701 | error = -EISDIR; | |
1702 | if (nd->last_type != LAST_NORM || nd->last.name[nd->last.len]) | |
1703 | goto exit; | |
1704 | ||
1705 | dir = nd->dentry; | |
1706 | nd->flags &= ~LOOKUP_PARENT; | |
1b1dcc1b | 1707 | mutex_lock(&dir->d_inode->i_mutex); |
49705b77 | 1708 | path.dentry = lookup_hash(nd); |
d73ffe16 | 1709 | path.mnt = nd->mnt; |
1da177e4 LT |
1710 | |
1711 | do_last: | |
4e7506e4 AV |
1712 | error = PTR_ERR(path.dentry); |
1713 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1714 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1715 | goto exit; |
1716 | } | |
1717 | ||
4af4c52f OD |
1718 | if (IS_ERR(nd->intent.open.file)) { |
1719 | mutex_unlock(&dir->d_inode->i_mutex); | |
1720 | error = PTR_ERR(nd->intent.open.file); | |
1721 | goto exit_dput; | |
1722 | } | |
1723 | ||
1da177e4 | 1724 | /* Negative dentry, just create the file */ |
4e7506e4 | 1725 | if (!path.dentry->d_inode) { |
aab520e2 | 1726 | error = open_namei_create(nd, &path, flag, mode); |
1da177e4 LT |
1727 | if (error) |
1728 | goto exit; | |
aab520e2 | 1729 | return 0; |
1da177e4 LT |
1730 | } |
1731 | ||
1732 | /* | |
1733 | * It already exists. | |
1734 | */ | |
1b1dcc1b | 1735 | mutex_unlock(&dir->d_inode->i_mutex); |
3e2efce0 | 1736 | audit_inode_update(path.dentry->d_inode); |
1da177e4 LT |
1737 | |
1738 | error = -EEXIST; | |
1739 | if (flag & O_EXCL) | |
1740 | goto exit_dput; | |
1741 | ||
e13b210f | 1742 | if (__follow_mount(&path)) { |
1da177e4 | 1743 | error = -ELOOP; |
ba7a4c1a AV |
1744 | if (flag & O_NOFOLLOW) |
1745 | goto exit_dput; | |
1da177e4 | 1746 | } |
3e2efce0 | 1747 | |
1da177e4 | 1748 | error = -ENOENT; |
4e7506e4 | 1749 | if (!path.dentry->d_inode) |
1da177e4 | 1750 | goto exit_dput; |
4e7506e4 | 1751 | if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1752 | goto do_link; |
1753 | ||
09dd17d3 | 1754 | path_to_nameidata(&path, nd); |
1da177e4 | 1755 | error = -EISDIR; |
4e7506e4 | 1756 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1757 | goto exit; |
1758 | ok: | |
1759 | error = may_open(nd, acc_mode, flag); | |
1760 | if (error) | |
1761 | goto exit; | |
1762 | return 0; | |
1763 | ||
1764 | exit_dput: | |
09dd17d3 | 1765 | dput_path(&path, nd); |
1da177e4 | 1766 | exit: |
834f2a4a TM |
1767 | if (!IS_ERR(nd->intent.open.file)) |
1768 | release_open_intent(nd); | |
1da177e4 LT |
1769 | path_release(nd); |
1770 | return error; | |
1771 | ||
1772 | do_link: | |
1773 | error = -ELOOP; | |
1774 | if (flag & O_NOFOLLOW) | |
1775 | goto exit_dput; | |
1776 | /* | |
1777 | * This is subtle. Instead of calling do_follow_link() we do the | |
1778 | * thing by hands. The reason is that this way we have zero link_count | |
1779 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1780 | * After that we have the parent and last component, i.e. | |
1781 | * we are in the same situation as after the first path_walk(). | |
1782 | * Well, almost - if the last component is normal we get its copy | |
1783 | * stored in nd->last.name and we will have to putname() it when we | |
1784 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1785 | */ | |
1786 | nd->flags |= LOOKUP_PARENT; | |
4e7506e4 | 1787 | error = security_inode_follow_link(path.dentry, nd); |
1da177e4 LT |
1788 | if (error) |
1789 | goto exit_dput; | |
cd4e91d3 | 1790 | error = __do_follow_link(&path, nd); |
de459215 KK |
1791 | if (error) { |
1792 | /* Does someone understand code flow here? Or it is only | |
1793 | * me so stupid? Anathema to whoever designed this non-sense | |
1794 | * with "intent.open". | |
1795 | */ | |
1796 | release_open_intent(nd); | |
1da177e4 | 1797 | return error; |
de459215 | 1798 | } |
1da177e4 | 1799 | nd->flags &= ~LOOKUP_PARENT; |
d671d5e5 | 1800 | if (nd->last_type == LAST_BIND) |
1da177e4 | 1801 | goto ok; |
1da177e4 LT |
1802 | error = -EISDIR; |
1803 | if (nd->last_type != LAST_NORM) | |
1804 | goto exit; | |
1805 | if (nd->last.name[nd->last.len]) { | |
82984114 | 1806 | __putname(nd->last.name); |
1da177e4 LT |
1807 | goto exit; |
1808 | } | |
1809 | error = -ELOOP; | |
1810 | if (count++==32) { | |
82984114 | 1811 | __putname(nd->last.name); |
1da177e4 LT |
1812 | goto exit; |
1813 | } | |
1814 | dir = nd->dentry; | |
1b1dcc1b | 1815 | mutex_lock(&dir->d_inode->i_mutex); |
49705b77 | 1816 | path.dentry = lookup_hash(nd); |
d671d5e5 | 1817 | path.mnt = nd->mnt; |
82984114 | 1818 | __putname(nd->last.name); |
1da177e4 LT |
1819 | goto do_last; |
1820 | } | |
1821 | ||
1822 | /** | |
1823 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1824 | * @nd: nameidata info | |
1825 | * @is_dir: directory flag | |
1826 | * | |
1827 | * Simple function to lookup and return a dentry and create it | |
1828 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1829 | * |
1b1dcc1b | 1830 | * Returns with nd->dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1831 | */ |
1832 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1833 | { | |
c663e5d8 | 1834 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1835 | |
f2eace23 | 1836 | mutex_lock_nested(&nd->dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1837 | /* |
1838 | * Yucky last component or no last component at all? | |
1839 | * (foo/., foo/.., /////) | |
1840 | */ | |
1da177e4 LT |
1841 | if (nd->last_type != LAST_NORM) |
1842 | goto fail; | |
1843 | nd->flags &= ~LOOKUP_PARENT; | |
a634904a AM |
1844 | nd->flags |= LOOKUP_CREATE; |
1845 | nd->intent.open.flags = O_EXCL; | |
c663e5d8 CH |
1846 | |
1847 | /* | |
1848 | * Do the final lookup. | |
1849 | */ | |
49705b77 | 1850 | dentry = lookup_hash(nd); |
1da177e4 LT |
1851 | if (IS_ERR(dentry)) |
1852 | goto fail; | |
c663e5d8 CH |
1853 | |
1854 | /* | |
1855 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1856 | * From the vfs_mknod() POV we just have a negative dentry - | |
1857 | * all is fine. Let's be bastards - you had / on the end, you've | |
1858 | * been asking for (non-existent) directory. -ENOENT for you. | |
1859 | */ | |
1da177e4 LT |
1860 | if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode) |
1861 | goto enoent; | |
1862 | return dentry; | |
1863 | enoent: | |
1864 | dput(dentry); | |
1865 | dentry = ERR_PTR(-ENOENT); | |
1866 | fail: | |
1867 | return dentry; | |
1868 | } | |
f81a0bff | 1869 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1870 | |
1871 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1872 | { | |
1873 | int error = may_create(dir, dentry, NULL); | |
1874 | ||
1875 | if (error) | |
1876 | return error; | |
1877 | ||
1878 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1879 | return -EPERM; | |
1880 | ||
1881 | if (!dir->i_op || !dir->i_op->mknod) | |
1882 | return -EPERM; | |
1883 | ||
1884 | error = security_inode_mknod(dir, dentry, mode, dev); | |
1885 | if (error) | |
1886 | return error; | |
1887 | ||
1888 | DQUOT_INIT(dir); | |
1889 | error = dir->i_op->mknod(dir, dentry, mode, dev); | |
a74574aa | 1890 | if (!error) |
f38aa942 | 1891 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1892 | return error; |
1893 | } | |
1894 | ||
5590ff0d UD |
1895 | asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode, |
1896 | unsigned dev) | |
1da177e4 LT |
1897 | { |
1898 | int error = 0; | |
1899 | char * tmp; | |
1900 | struct dentry * dentry; | |
1901 | struct nameidata nd; | |
1902 | ||
1903 | if (S_ISDIR(mode)) | |
1904 | return -EPERM; | |
1905 | tmp = getname(filename); | |
1906 | if (IS_ERR(tmp)) | |
1907 | return PTR_ERR(tmp); | |
1908 | ||
5590ff0d | 1909 | error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd); |
1da177e4 LT |
1910 | if (error) |
1911 | goto out; | |
1912 | dentry = lookup_create(&nd, 0); | |
1913 | error = PTR_ERR(dentry); | |
1914 | ||
1915 | if (!IS_POSIXACL(nd.dentry->d_inode)) | |
1916 | mode &= ~current->fs->umask; | |
1917 | if (!IS_ERR(dentry)) { | |
1918 | switch (mode & S_IFMT) { | |
1919 | case 0: case S_IFREG: | |
1920 | error = vfs_create(nd.dentry->d_inode,dentry,mode,&nd); | |
1921 | break; | |
1922 | case S_IFCHR: case S_IFBLK: | |
1923 | error = vfs_mknod(nd.dentry->d_inode,dentry,mode, | |
1924 | new_decode_dev(dev)); | |
1925 | break; | |
1926 | case S_IFIFO: case S_IFSOCK: | |
1927 | error = vfs_mknod(nd.dentry->d_inode,dentry,mode,0); | |
1928 | break; | |
1929 | case S_IFDIR: | |
1930 | error = -EPERM; | |
1931 | break; | |
1932 | default: | |
1933 | error = -EINVAL; | |
1934 | } | |
1935 | dput(dentry); | |
1936 | } | |
1b1dcc1b | 1937 | mutex_unlock(&nd.dentry->d_inode->i_mutex); |
1da177e4 LT |
1938 | path_release(&nd); |
1939 | out: | |
1940 | putname(tmp); | |
1941 | ||
1942 | return error; | |
1943 | } | |
1944 | ||
5590ff0d UD |
1945 | asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev) |
1946 | { | |
1947 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
1948 | } | |
1949 | ||
1da177e4 LT |
1950 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
1951 | { | |
1952 | int error = may_create(dir, dentry, NULL); | |
1953 | ||
1954 | if (error) | |
1955 | return error; | |
1956 | ||
1957 | if (!dir->i_op || !dir->i_op->mkdir) | |
1958 | return -EPERM; | |
1959 | ||
1960 | mode &= (S_IRWXUGO|S_ISVTX); | |
1961 | error = security_inode_mkdir(dir, dentry, mode); | |
1962 | if (error) | |
1963 | return error; | |
1964 | ||
1965 | DQUOT_INIT(dir); | |
1966 | error = dir->i_op->mkdir(dir, dentry, mode); | |
a74574aa | 1967 | if (!error) |
f38aa942 | 1968 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
1969 | return error; |
1970 | } | |
1971 | ||
5590ff0d | 1972 | asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode) |
1da177e4 LT |
1973 | { |
1974 | int error = 0; | |
1975 | char * tmp; | |
6902d925 DH |
1976 | struct dentry *dentry; |
1977 | struct nameidata nd; | |
1da177e4 LT |
1978 | |
1979 | tmp = getname(pathname); | |
1980 | error = PTR_ERR(tmp); | |
6902d925 DH |
1981 | if (IS_ERR(tmp)) |
1982 | goto out_err; | |
1da177e4 | 1983 | |
6902d925 DH |
1984 | error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd); |
1985 | if (error) | |
1986 | goto out; | |
1987 | dentry = lookup_create(&nd, 1); | |
1988 | error = PTR_ERR(dentry); | |
1989 | if (IS_ERR(dentry)) | |
1990 | goto out_unlock; | |
1da177e4 | 1991 | |
6902d925 DH |
1992 | if (!IS_POSIXACL(nd.dentry->d_inode)) |
1993 | mode &= ~current->fs->umask; | |
1994 | error = vfs_mkdir(nd.dentry->d_inode, dentry, mode); | |
1995 | dput(dentry); | |
1996 | out_unlock: | |
1997 | mutex_unlock(&nd.dentry->d_inode->i_mutex); | |
1998 | path_release(&nd); | |
1999 | out: | |
2000 | putname(tmp); | |
2001 | out_err: | |
1da177e4 LT |
2002 | return error; |
2003 | } | |
2004 | ||
5590ff0d UD |
2005 | asmlinkage long sys_mkdir(const char __user *pathname, int mode) |
2006 | { | |
2007 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2008 | } | |
2009 | ||
1da177e4 LT |
2010 | /* |
2011 | * We try to drop the dentry early: we should have | |
2012 | * a usage count of 2 if we're the only user of this | |
2013 | * dentry, and if that is true (possibly after pruning | |
2014 | * the dcache), then we drop the dentry now. | |
2015 | * | |
2016 | * A low-level filesystem can, if it choses, legally | |
2017 | * do a | |
2018 | * | |
2019 | * if (!d_unhashed(dentry)) | |
2020 | * return -EBUSY; | |
2021 | * | |
2022 | * if it cannot handle the case of removing a directory | |
2023 | * that is still in use by something else.. | |
2024 | */ | |
2025 | void dentry_unhash(struct dentry *dentry) | |
2026 | { | |
2027 | dget(dentry); | |
dc168427 | 2028 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2029 | spin_lock(&dcache_lock); |
2030 | spin_lock(&dentry->d_lock); | |
2031 | if (atomic_read(&dentry->d_count) == 2) | |
2032 | __d_drop(dentry); | |
2033 | spin_unlock(&dentry->d_lock); | |
2034 | spin_unlock(&dcache_lock); | |
2035 | } | |
2036 | ||
2037 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2038 | { | |
2039 | int error = may_delete(dir, dentry, 1); | |
2040 | ||
2041 | if (error) | |
2042 | return error; | |
2043 | ||
2044 | if (!dir->i_op || !dir->i_op->rmdir) | |
2045 | return -EPERM; | |
2046 | ||
2047 | DQUOT_INIT(dir); | |
2048 | ||
1b1dcc1b | 2049 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2050 | dentry_unhash(dentry); |
2051 | if (d_mountpoint(dentry)) | |
2052 | error = -EBUSY; | |
2053 | else { | |
2054 | error = security_inode_rmdir(dir, dentry); | |
2055 | if (!error) { | |
2056 | error = dir->i_op->rmdir(dir, dentry); | |
2057 | if (!error) | |
2058 | dentry->d_inode->i_flags |= S_DEAD; | |
2059 | } | |
2060 | } | |
1b1dcc1b | 2061 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2062 | if (!error) { |
1da177e4 LT |
2063 | d_delete(dentry); |
2064 | } | |
2065 | dput(dentry); | |
2066 | ||
2067 | return error; | |
2068 | } | |
2069 | ||
5590ff0d | 2070 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2071 | { |
2072 | int error = 0; | |
2073 | char * name; | |
2074 | struct dentry *dentry; | |
2075 | struct nameidata nd; | |
2076 | ||
2077 | name = getname(pathname); | |
2078 | if(IS_ERR(name)) | |
2079 | return PTR_ERR(name); | |
2080 | ||
5590ff0d | 2081 | error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2082 | if (error) |
2083 | goto exit; | |
2084 | ||
2085 | switch(nd.last_type) { | |
2086 | case LAST_DOTDOT: | |
2087 | error = -ENOTEMPTY; | |
2088 | goto exit1; | |
2089 | case LAST_DOT: | |
2090 | error = -EINVAL; | |
2091 | goto exit1; | |
2092 | case LAST_ROOT: | |
2093 | error = -EBUSY; | |
2094 | goto exit1; | |
2095 | } | |
f2eace23 | 2096 | mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2097 | dentry = lookup_hash(&nd); |
1da177e4 | 2098 | error = PTR_ERR(dentry); |
6902d925 DH |
2099 | if (IS_ERR(dentry)) |
2100 | goto exit2; | |
2101 | error = vfs_rmdir(nd.dentry->d_inode, dentry); | |
2102 | dput(dentry); | |
2103 | exit2: | |
1b1dcc1b | 2104 | mutex_unlock(&nd.dentry->d_inode->i_mutex); |
1da177e4 LT |
2105 | exit1: |
2106 | path_release(&nd); | |
2107 | exit: | |
2108 | putname(name); | |
2109 | return error; | |
2110 | } | |
2111 | ||
5590ff0d UD |
2112 | asmlinkage long sys_rmdir(const char __user *pathname) |
2113 | { | |
2114 | return do_rmdir(AT_FDCWD, pathname); | |
2115 | } | |
2116 | ||
1da177e4 LT |
2117 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2118 | { | |
2119 | int error = may_delete(dir, dentry, 0); | |
2120 | ||
2121 | if (error) | |
2122 | return error; | |
2123 | ||
2124 | if (!dir->i_op || !dir->i_op->unlink) | |
2125 | return -EPERM; | |
2126 | ||
2127 | DQUOT_INIT(dir); | |
2128 | ||
1b1dcc1b | 2129 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2130 | if (d_mountpoint(dentry)) |
2131 | error = -EBUSY; | |
2132 | else { | |
2133 | error = security_inode_unlink(dir, dentry); | |
2134 | if (!error) | |
2135 | error = dir->i_op->unlink(dir, dentry); | |
2136 | } | |
1b1dcc1b | 2137 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2138 | |
2139 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2140 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
e234f35c | 2141 | d_delete(dentry); |
1da177e4 | 2142 | } |
0eeca283 | 2143 | |
1da177e4 LT |
2144 | return error; |
2145 | } | |
2146 | ||
2147 | /* | |
2148 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2149 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2150 | * writeout happening, and we don't want to prevent access to the directory |
2151 | * while waiting on the I/O. | |
2152 | */ | |
5590ff0d | 2153 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 LT |
2154 | { |
2155 | int error = 0; | |
2156 | char * name; | |
2157 | struct dentry *dentry; | |
2158 | struct nameidata nd; | |
2159 | struct inode *inode = NULL; | |
2160 | ||
2161 | name = getname(pathname); | |
2162 | if(IS_ERR(name)) | |
2163 | return PTR_ERR(name); | |
2164 | ||
5590ff0d | 2165 | error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2166 | if (error) |
2167 | goto exit; | |
2168 | error = -EISDIR; | |
2169 | if (nd.last_type != LAST_NORM) | |
2170 | goto exit1; | |
f2eace23 | 2171 | mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2172 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2173 | error = PTR_ERR(dentry); |
2174 | if (!IS_ERR(dentry)) { | |
2175 | /* Why not before? Because we want correct error value */ | |
2176 | if (nd.last.name[nd.last.len]) | |
2177 | goto slashes; | |
2178 | inode = dentry->d_inode; | |
2179 | if (inode) | |
2180 | atomic_inc(&inode->i_count); | |
2181 | error = vfs_unlink(nd.dentry->d_inode, dentry); | |
2182 | exit2: | |
2183 | dput(dentry); | |
2184 | } | |
1b1dcc1b | 2185 | mutex_unlock(&nd.dentry->d_inode->i_mutex); |
1da177e4 LT |
2186 | if (inode) |
2187 | iput(inode); /* truncate the inode here */ | |
2188 | exit1: | |
2189 | path_release(&nd); | |
2190 | exit: | |
2191 | putname(name); | |
2192 | return error; | |
2193 | ||
2194 | slashes: | |
2195 | error = !dentry->d_inode ? -ENOENT : | |
2196 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2197 | goto exit2; | |
2198 | } | |
2199 | ||
5590ff0d UD |
2200 | asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag) |
2201 | { | |
2202 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2203 | return -EINVAL; | |
2204 | ||
2205 | if (flag & AT_REMOVEDIR) | |
2206 | return do_rmdir(dfd, pathname); | |
2207 | ||
2208 | return do_unlinkat(dfd, pathname); | |
2209 | } | |
2210 | ||
2211 | asmlinkage long sys_unlink(const char __user *pathname) | |
2212 | { | |
2213 | return do_unlinkat(AT_FDCWD, pathname); | |
2214 | } | |
2215 | ||
1da177e4 LT |
2216 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname, int mode) |
2217 | { | |
2218 | int error = may_create(dir, dentry, NULL); | |
2219 | ||
2220 | if (error) | |
2221 | return error; | |
2222 | ||
2223 | if (!dir->i_op || !dir->i_op->symlink) | |
2224 | return -EPERM; | |
2225 | ||
2226 | error = security_inode_symlink(dir, dentry, oldname); | |
2227 | if (error) | |
2228 | return error; | |
2229 | ||
2230 | DQUOT_INIT(dir); | |
2231 | error = dir->i_op->symlink(dir, dentry, oldname); | |
a74574aa | 2232 | if (!error) |
f38aa942 | 2233 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2234 | return error; |
2235 | } | |
2236 | ||
5590ff0d UD |
2237 | asmlinkage long sys_symlinkat(const char __user *oldname, |
2238 | int newdfd, const char __user *newname) | |
1da177e4 LT |
2239 | { |
2240 | int error = 0; | |
2241 | char * from; | |
2242 | char * to; | |
6902d925 DH |
2243 | struct dentry *dentry; |
2244 | struct nameidata nd; | |
1da177e4 LT |
2245 | |
2246 | from = getname(oldname); | |
2247 | if(IS_ERR(from)) | |
2248 | return PTR_ERR(from); | |
2249 | to = getname(newname); | |
2250 | error = PTR_ERR(to); | |
6902d925 DH |
2251 | if (IS_ERR(to)) |
2252 | goto out_putname; | |
1da177e4 | 2253 | |
6902d925 DH |
2254 | error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd); |
2255 | if (error) | |
2256 | goto out; | |
2257 | dentry = lookup_create(&nd, 0); | |
2258 | error = PTR_ERR(dentry); | |
2259 | if (IS_ERR(dentry)) | |
2260 | goto out_unlock; | |
2261 | ||
2262 | error = vfs_symlink(nd.dentry->d_inode, dentry, from, S_IALLUGO); | |
2263 | dput(dentry); | |
2264 | out_unlock: | |
2265 | mutex_unlock(&nd.dentry->d_inode->i_mutex); | |
2266 | path_release(&nd); | |
1da177e4 | 2267 | out: |
6902d925 DH |
2268 | putname(to); |
2269 | out_putname: | |
1da177e4 LT |
2270 | putname(from); |
2271 | return error; | |
2272 | } | |
2273 | ||
5590ff0d UD |
2274 | asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname) |
2275 | { | |
2276 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2277 | } | |
2278 | ||
1da177e4 LT |
2279 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2280 | { | |
2281 | struct inode *inode = old_dentry->d_inode; | |
2282 | int error; | |
2283 | ||
2284 | if (!inode) | |
2285 | return -ENOENT; | |
2286 | ||
2287 | error = may_create(dir, new_dentry, NULL); | |
2288 | if (error) | |
2289 | return error; | |
2290 | ||
2291 | if (dir->i_sb != inode->i_sb) | |
2292 | return -EXDEV; | |
2293 | ||
2294 | /* | |
2295 | * A link to an append-only or immutable file cannot be created. | |
2296 | */ | |
2297 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2298 | return -EPERM; | |
2299 | if (!dir->i_op || !dir->i_op->link) | |
2300 | return -EPERM; | |
2301 | if (S_ISDIR(old_dentry->d_inode->i_mode)) | |
2302 | return -EPERM; | |
2303 | ||
2304 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2305 | if (error) | |
2306 | return error; | |
2307 | ||
1b1dcc1b | 2308 | mutex_lock(&old_dentry->d_inode->i_mutex); |
1da177e4 LT |
2309 | DQUOT_INIT(dir); |
2310 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
1b1dcc1b | 2311 | mutex_unlock(&old_dentry->d_inode->i_mutex); |
e31e14ec | 2312 | if (!error) |
f38aa942 | 2313 | fsnotify_create(dir, new_dentry); |
1da177e4 LT |
2314 | return error; |
2315 | } | |
2316 | ||
2317 | /* | |
2318 | * Hardlinks are often used in delicate situations. We avoid | |
2319 | * security-related surprises by not following symlinks on the | |
2320 | * newname. --KAB | |
2321 | * | |
2322 | * We don't follow them on the oldname either to be compatible | |
2323 | * with linux 2.0, and to avoid hard-linking to directories | |
2324 | * and other special files. --ADM | |
2325 | */ | |
5590ff0d | 2326 | asmlinkage long sys_linkat(int olddfd, const char __user *oldname, |
c04030e1 UD |
2327 | int newdfd, const char __user *newname, |
2328 | int flags) | |
1da177e4 LT |
2329 | { |
2330 | struct dentry *new_dentry; | |
2331 | struct nameidata nd, old_nd; | |
2332 | int error; | |
2333 | char * to; | |
2334 | ||
45c9b11a | 2335 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2336 | return -EINVAL; |
2337 | ||
1da177e4 LT |
2338 | to = getname(newname); |
2339 | if (IS_ERR(to)) | |
2340 | return PTR_ERR(to); | |
2341 | ||
45c9b11a UD |
2342 | error = __user_walk_fd(olddfd, oldname, |
2343 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2344 | &old_nd); | |
1da177e4 LT |
2345 | if (error) |
2346 | goto exit; | |
5590ff0d | 2347 | error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd); |
1da177e4 LT |
2348 | if (error) |
2349 | goto out; | |
2350 | error = -EXDEV; | |
2351 | if (old_nd.mnt != nd.mnt) | |
2352 | goto out_release; | |
2353 | new_dentry = lookup_create(&nd, 0); | |
2354 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2355 | if (IS_ERR(new_dentry)) |
2356 | goto out_unlock; | |
2357 | error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry); | |
2358 | dput(new_dentry); | |
2359 | out_unlock: | |
1b1dcc1b | 2360 | mutex_unlock(&nd.dentry->d_inode->i_mutex); |
1da177e4 LT |
2361 | out_release: |
2362 | path_release(&nd); | |
2363 | out: | |
2364 | path_release(&old_nd); | |
2365 | exit: | |
2366 | putname(to); | |
2367 | ||
2368 | return error; | |
2369 | } | |
2370 | ||
5590ff0d UD |
2371 | asmlinkage long sys_link(const char __user *oldname, const char __user *newname) |
2372 | { | |
c04030e1 | 2373 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2374 | } |
2375 | ||
1da177e4 LT |
2376 | /* |
2377 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2378 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2379 | * Problems: | |
2380 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2381 | * b) race potential - two innocent renames can create a loop together. | |
2382 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2383 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2384 | * story. |
2385 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2386 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2387 | * whether the target exists). Solution: try to be smart with locking |
2388 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2389 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2390 | * move will be locked. Thus we can rank directories by the tree |
2391 | * (ancestors first) and rank all non-directories after them. | |
2392 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2393 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2394 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2395 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2396 | * we'd better make sure that there's no link(2) for them. | |
2397 | * d) some filesystems don't support opened-but-unlinked directories, | |
2398 | * either because of layout or because they are not ready to deal with | |
2399 | * all cases correctly. The latter will be fixed (taking this sort of | |
2400 | * stuff into VFS), but the former is not going away. Solution: the same | |
2401 | * trick as in rmdir(). | |
2402 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2403 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2404 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2405 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2406 | * locking]. |
2407 | */ | |
75c96f85 AB |
2408 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2409 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2410 | { |
2411 | int error = 0; | |
2412 | struct inode *target; | |
2413 | ||
2414 | /* | |
2415 | * If we are going to change the parent - check write permissions, | |
2416 | * we'll need to flip '..'. | |
2417 | */ | |
2418 | if (new_dir != old_dir) { | |
2419 | error = permission(old_dentry->d_inode, MAY_WRITE, NULL); | |
2420 | if (error) | |
2421 | return error; | |
2422 | } | |
2423 | ||
2424 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2425 | if (error) | |
2426 | return error; | |
2427 | ||
2428 | target = new_dentry->d_inode; | |
2429 | if (target) { | |
1b1dcc1b | 2430 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2431 | dentry_unhash(new_dentry); |
2432 | } | |
2433 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2434 | error = -EBUSY; | |
2435 | else | |
2436 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2437 | if (target) { | |
2438 | if (!error) | |
2439 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2440 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2441 | if (d_unhashed(new_dentry)) |
2442 | d_rehash(new_dentry); | |
2443 | dput(new_dentry); | |
2444 | } | |
e31e14ec | 2445 | if (!error) |
349457cc MF |
2446 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2447 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2448 | return error; |
2449 | } | |
2450 | ||
75c96f85 AB |
2451 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2452 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2453 | { |
2454 | struct inode *target; | |
2455 | int error; | |
2456 | ||
2457 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2458 | if (error) | |
2459 | return error; | |
2460 | ||
2461 | dget(new_dentry); | |
2462 | target = new_dentry->d_inode; | |
2463 | if (target) | |
1b1dcc1b | 2464 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2465 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2466 | error = -EBUSY; | |
2467 | else | |
2468 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2469 | if (!error) { | |
349457cc | 2470 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2471 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2472 | } |
2473 | if (target) | |
1b1dcc1b | 2474 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2475 | dput(new_dentry); |
2476 | return error; | |
2477 | } | |
2478 | ||
2479 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2480 | struct inode *new_dir, struct dentry *new_dentry) | |
2481 | { | |
2482 | int error; | |
2483 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2484 | const char *old_name; |
1da177e4 LT |
2485 | |
2486 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2487 | return 0; | |
2488 | ||
2489 | error = may_delete(old_dir, old_dentry, is_dir); | |
2490 | if (error) | |
2491 | return error; | |
2492 | ||
2493 | if (!new_dentry->d_inode) | |
2494 | error = may_create(new_dir, new_dentry, NULL); | |
2495 | else | |
2496 | error = may_delete(new_dir, new_dentry, is_dir); | |
2497 | if (error) | |
2498 | return error; | |
2499 | ||
2500 | if (!old_dir->i_op || !old_dir->i_op->rename) | |
2501 | return -EPERM; | |
2502 | ||
2503 | DQUOT_INIT(old_dir); | |
2504 | DQUOT_INIT(new_dir); | |
2505 | ||
0eeca283 RL |
2506 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2507 | ||
1da177e4 LT |
2508 | if (is_dir) |
2509 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2510 | else | |
2511 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2512 | if (!error) { | |
0eeca283 | 2513 | const char *new_name = old_dentry->d_name.name; |
89204c40 JM |
2514 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
2515 | new_dentry->d_inode, old_dentry->d_inode); | |
1da177e4 | 2516 | } |
0eeca283 RL |
2517 | fsnotify_oldname_free(old_name); |
2518 | ||
1da177e4 LT |
2519 | return error; |
2520 | } | |
2521 | ||
5590ff0d UD |
2522 | static int do_rename(int olddfd, const char *oldname, |
2523 | int newdfd, const char *newname) | |
1da177e4 LT |
2524 | { |
2525 | int error = 0; | |
2526 | struct dentry * old_dir, * new_dir; | |
2527 | struct dentry * old_dentry, *new_dentry; | |
2528 | struct dentry * trap; | |
2529 | struct nameidata oldnd, newnd; | |
2530 | ||
5590ff0d | 2531 | error = do_path_lookup(olddfd, oldname, LOOKUP_PARENT, &oldnd); |
1da177e4 LT |
2532 | if (error) |
2533 | goto exit; | |
2534 | ||
5590ff0d | 2535 | error = do_path_lookup(newdfd, newname, LOOKUP_PARENT, &newnd); |
1da177e4 LT |
2536 | if (error) |
2537 | goto exit1; | |
2538 | ||
2539 | error = -EXDEV; | |
2540 | if (oldnd.mnt != newnd.mnt) | |
2541 | goto exit2; | |
2542 | ||
2543 | old_dir = oldnd.dentry; | |
2544 | error = -EBUSY; | |
2545 | if (oldnd.last_type != LAST_NORM) | |
2546 | goto exit2; | |
2547 | ||
2548 | new_dir = newnd.dentry; | |
2549 | if (newnd.last_type != LAST_NORM) | |
2550 | goto exit2; | |
2551 | ||
2552 | trap = lock_rename(new_dir, old_dir); | |
2553 | ||
49705b77 | 2554 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2555 | error = PTR_ERR(old_dentry); |
2556 | if (IS_ERR(old_dentry)) | |
2557 | goto exit3; | |
2558 | /* source must exist */ | |
2559 | error = -ENOENT; | |
2560 | if (!old_dentry->d_inode) | |
2561 | goto exit4; | |
2562 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2563 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2564 | error = -ENOTDIR; | |
2565 | if (oldnd.last.name[oldnd.last.len]) | |
2566 | goto exit4; | |
2567 | if (newnd.last.name[newnd.last.len]) | |
2568 | goto exit4; | |
2569 | } | |
2570 | /* source should not be ancestor of target */ | |
2571 | error = -EINVAL; | |
2572 | if (old_dentry == trap) | |
2573 | goto exit4; | |
49705b77 | 2574 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2575 | error = PTR_ERR(new_dentry); |
2576 | if (IS_ERR(new_dentry)) | |
2577 | goto exit4; | |
2578 | /* target should not be an ancestor of source */ | |
2579 | error = -ENOTEMPTY; | |
2580 | if (new_dentry == trap) | |
2581 | goto exit5; | |
2582 | ||
2583 | error = vfs_rename(old_dir->d_inode, old_dentry, | |
2584 | new_dir->d_inode, new_dentry); | |
2585 | exit5: | |
2586 | dput(new_dentry); | |
2587 | exit4: | |
2588 | dput(old_dentry); | |
2589 | exit3: | |
2590 | unlock_rename(new_dir, old_dir); | |
2591 | exit2: | |
2592 | path_release(&newnd); | |
2593 | exit1: | |
2594 | path_release(&oldnd); | |
2595 | exit: | |
2596 | return error; | |
2597 | } | |
2598 | ||
5590ff0d UD |
2599 | asmlinkage long sys_renameat(int olddfd, const char __user *oldname, |
2600 | int newdfd, const char __user *newname) | |
1da177e4 LT |
2601 | { |
2602 | int error; | |
2603 | char * from; | |
2604 | char * to; | |
2605 | ||
2606 | from = getname(oldname); | |
2607 | if(IS_ERR(from)) | |
2608 | return PTR_ERR(from); | |
2609 | to = getname(newname); | |
2610 | error = PTR_ERR(to); | |
2611 | if (!IS_ERR(to)) { | |
5590ff0d | 2612 | error = do_rename(olddfd, from, newdfd, to); |
1da177e4 LT |
2613 | putname(to); |
2614 | } | |
2615 | putname(from); | |
2616 | return error; | |
2617 | } | |
2618 | ||
5590ff0d UD |
2619 | asmlinkage long sys_rename(const char __user *oldname, const char __user *newname) |
2620 | { | |
2621 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2622 | } | |
2623 | ||
1da177e4 LT |
2624 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2625 | { | |
2626 | int len; | |
2627 | ||
2628 | len = PTR_ERR(link); | |
2629 | if (IS_ERR(link)) | |
2630 | goto out; | |
2631 | ||
2632 | len = strlen(link); | |
2633 | if (len > (unsigned) buflen) | |
2634 | len = buflen; | |
2635 | if (copy_to_user(buffer, link, len)) | |
2636 | len = -EFAULT; | |
2637 | out: | |
2638 | return len; | |
2639 | } | |
2640 | ||
2641 | /* | |
2642 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2643 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2644 | * using) it for any given inode is up to filesystem. | |
2645 | */ | |
2646 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2647 | { | |
2648 | struct nameidata nd; | |
cc314eef LT |
2649 | void *cookie; |
2650 | ||
1da177e4 | 2651 | nd.depth = 0; |
cc314eef LT |
2652 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
2653 | if (!IS_ERR(cookie)) { | |
2654 | int res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
1da177e4 | 2655 | if (dentry->d_inode->i_op->put_link) |
cc314eef LT |
2656 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); |
2657 | cookie = ERR_PTR(res); | |
1da177e4 | 2658 | } |
cc314eef | 2659 | return PTR_ERR(cookie); |
1da177e4 LT |
2660 | } |
2661 | ||
2662 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2663 | { | |
2664 | return __vfs_follow_link(nd, link); | |
2665 | } | |
2666 | ||
2667 | /* get the link contents into pagecache */ | |
2668 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2669 | { | |
2670 | struct page * page; | |
2671 | struct address_space *mapping = dentry->d_inode->i_mapping; | |
090d2b18 | 2672 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 LT |
2673 | if (IS_ERR(page)) |
2674 | goto sync_fail; | |
2675 | wait_on_page_locked(page); | |
2676 | if (!PageUptodate(page)) | |
2677 | goto async_fail; | |
2678 | *ppage = page; | |
2679 | return kmap(page); | |
2680 | ||
2681 | async_fail: | |
2682 | page_cache_release(page); | |
2683 | return ERR_PTR(-EIO); | |
2684 | ||
2685 | sync_fail: | |
2686 | return (char*)page; | |
2687 | } | |
2688 | ||
2689 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2690 | { | |
2691 | struct page *page = NULL; | |
2692 | char *s = page_getlink(dentry, &page); | |
2693 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2694 | if (page) { | |
2695 | kunmap(page); | |
2696 | page_cache_release(page); | |
2697 | } | |
2698 | return res; | |
2699 | } | |
2700 | ||
cc314eef | 2701 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2702 | { |
cc314eef | 2703 | struct page *page = NULL; |
1da177e4 | 2704 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2705 | return page; |
1da177e4 LT |
2706 | } |
2707 | ||
cc314eef | 2708 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2709 | { |
cc314eef LT |
2710 | struct page *page = cookie; |
2711 | ||
2712 | if (page) { | |
1da177e4 LT |
2713 | kunmap(page); |
2714 | page_cache_release(page); | |
1da177e4 LT |
2715 | } |
2716 | } | |
2717 | ||
0adb25d2 KK |
2718 | int __page_symlink(struct inode *inode, const char *symname, int len, |
2719 | gfp_t gfp_mask) | |
1da177e4 LT |
2720 | { |
2721 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2722 | struct page *page; |
beb497ab | 2723 | int err; |
1da177e4 LT |
2724 | char *kaddr; |
2725 | ||
7e53cac4 | 2726 | retry: |
beb497ab | 2727 | err = -ENOMEM; |
0adb25d2 | 2728 | page = find_or_create_page(mapping, 0, gfp_mask); |
1da177e4 LT |
2729 | if (!page) |
2730 | goto fail; | |
2731 | err = mapping->a_ops->prepare_write(NULL, page, 0, len-1); | |
7e53cac4 N |
2732 | if (err == AOP_TRUNCATED_PAGE) { |
2733 | page_cache_release(page); | |
2734 | goto retry; | |
2735 | } | |
1da177e4 LT |
2736 | if (err) |
2737 | goto fail_map; | |
2738 | kaddr = kmap_atomic(page, KM_USER0); | |
2739 | memcpy(kaddr, symname, len-1); | |
2740 | kunmap_atomic(kaddr, KM_USER0); | |
7e53cac4 N |
2741 | err = mapping->a_ops->commit_write(NULL, page, 0, len-1); |
2742 | if (err == AOP_TRUNCATED_PAGE) { | |
2743 | page_cache_release(page); | |
2744 | goto retry; | |
2745 | } | |
2746 | if (err) | |
2747 | goto fail_map; | |
1da177e4 LT |
2748 | /* |
2749 | * Notice that we are _not_ going to block here - end of page is | |
2750 | * unmapped, so this will only try to map the rest of page, see | |
2751 | * that it is unmapped (typically even will not look into inode - | |
2752 | * ->i_size will be enough for everything) and zero it out. | |
2753 | * OTOH it's obviously correct and should make the page up-to-date. | |
2754 | */ | |
2755 | if (!PageUptodate(page)) { | |
2756 | err = mapping->a_ops->readpage(NULL, page); | |
7e53cac4 N |
2757 | if (err != AOP_TRUNCATED_PAGE) |
2758 | wait_on_page_locked(page); | |
1da177e4 LT |
2759 | } else { |
2760 | unlock_page(page); | |
2761 | } | |
2762 | page_cache_release(page); | |
2763 | if (err < 0) | |
2764 | goto fail; | |
2765 | mark_inode_dirty(inode); | |
2766 | return 0; | |
2767 | fail_map: | |
2768 | unlock_page(page); | |
2769 | page_cache_release(page); | |
2770 | fail: | |
2771 | return err; | |
2772 | } | |
2773 | ||
0adb25d2 KK |
2774 | int page_symlink(struct inode *inode, const char *symname, int len) |
2775 | { | |
2776 | return __page_symlink(inode, symname, len, | |
2777 | mapping_gfp_mask(inode->i_mapping)); | |
2778 | } | |
2779 | ||
92e1d5be | 2780 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2781 | .readlink = generic_readlink, |
2782 | .follow_link = page_follow_link_light, | |
2783 | .put_link = page_put_link, | |
2784 | }; | |
2785 | ||
2786 | EXPORT_SYMBOL(__user_walk); | |
5590ff0d | 2787 | EXPORT_SYMBOL(__user_walk_fd); |
1da177e4 LT |
2788 | EXPORT_SYMBOL(follow_down); |
2789 | EXPORT_SYMBOL(follow_up); | |
2790 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2791 | EXPORT_SYMBOL(getname); | |
2792 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2793 | EXPORT_SYMBOL(lookup_one_len); |
2794 | EXPORT_SYMBOL(page_follow_link_light); | |
2795 | EXPORT_SYMBOL(page_put_link); | |
2796 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2797 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2798 | EXPORT_SYMBOL(page_symlink); |
2799 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2800 | EXPORT_SYMBOL(path_lookup); | |
2801 | EXPORT_SYMBOL(path_release); | |
2802 | EXPORT_SYMBOL(path_walk); | |
2803 | EXPORT_SYMBOL(permission); | |
e4543edd | 2804 | EXPORT_SYMBOL(vfs_permission); |
8c744fb8 | 2805 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2806 | EXPORT_SYMBOL(unlock_rename); |
2807 | EXPORT_SYMBOL(vfs_create); | |
2808 | EXPORT_SYMBOL(vfs_follow_link); | |
2809 | EXPORT_SYMBOL(vfs_link); | |
2810 | EXPORT_SYMBOL(vfs_mkdir); | |
2811 | EXPORT_SYMBOL(vfs_mknod); | |
2812 | EXPORT_SYMBOL(generic_permission); | |
2813 | EXPORT_SYMBOL(vfs_readlink); | |
2814 | EXPORT_SYMBOL(vfs_rename); | |
2815 | EXPORT_SYMBOL(vfs_rmdir); | |
2816 | EXPORT_SYMBOL(vfs_symlink); | |
2817 | EXPORT_SYMBOL(vfs_unlink); | |
2818 | EXPORT_SYMBOL(dentry_unhash); | |
2819 | EXPORT_SYMBOL(generic_readlink); |