]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/fs.h> | |
21 | #include <linux/namei.h> | |
22 | #include <linux/quotaops.h> | |
23 | #include <linux/pagemap.h> | |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
6146f0d5 | 27 | #include <linux/ima.h> |
1da177e4 LT |
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 | 33 | #include <linux/fcntl.h> |
08ce5f16 | 34 | #include <linux/device_cgroup.h> |
1da177e4 LT |
35 | #include <asm/uaccess.h> |
36 | ||
37 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
38 | ||
39 | /* [Feb-1997 T. Schoebel-Theuer] | |
40 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
41 | * were necessary because of omirr. The reason is that omirr needs | |
42 | * to know the _real_ pathname, not the user-supplied one, in case | |
43 | * of symlinks (and also when transname replacements occur). | |
44 | * | |
45 | * The new code replaces the old recursive symlink resolution with | |
46 | * an iterative one (in case of non-nested symlink chains). It does | |
47 | * this with calls to <fs>_follow_link(). | |
48 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
49 | * replaced with a single function lookup_dentry() that can handle all | |
50 | * the special cases of the former code. | |
51 | * | |
52 | * With the new dcache, the pathname is stored at each inode, at least as | |
53 | * long as the refcount of the inode is positive. As a side effect, the | |
54 | * size of the dcache depends on the inode cache and thus is dynamic. | |
55 | * | |
56 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
57 | * resolution to correspond with current state of the code. | |
58 | * | |
59 | * Note that the symlink resolution is not *completely* iterative. | |
60 | * There is still a significant amount of tail- and mid- recursion in | |
61 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
62 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
63 | * may return different results than <fs>_follow_link(). Many virtual | |
64 | * filesystems (including /proc) exhibit this behavior. | |
65 | */ | |
66 | ||
67 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
68 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
69 | * and the name already exists in form of a symlink, try to create the new | |
70 | * name indicated by the symlink. The old code always complained that the | |
71 | * name already exists, due to not following the symlink even if its target | |
72 | * is nonexistent. The new semantics affects also mknod() and link() when | |
73 | * the name is a symlink pointing to a non-existant name. | |
74 | * | |
75 | * I don't know which semantics is the right one, since I have no access | |
76 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
77 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
78 | * "old" one. Personally, I think the new semantics is much more logical. | |
79 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
80 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
81 | * and in the old Linux semantics. | |
82 | */ | |
83 | ||
84 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
85 | * semantics. See the comments in "open_namei" and "do_link" below. | |
86 | * | |
87 | * [10-Sep-98 Alan Modra] Another symlink change. | |
88 | */ | |
89 | ||
90 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
91 | * inside the path - always follow. | |
92 | * in the last component in creation/removal/renaming - never follow. | |
93 | * if LOOKUP_FOLLOW passed - follow. | |
94 | * if the pathname has trailing slashes - follow. | |
95 | * otherwise - don't follow. | |
96 | * (applied in that order). | |
97 | * | |
98 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
99 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
100 | * During the 2.4 we need to fix the userland stuff depending on it - | |
101 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
102 | * XEmacs seems to be relying on it... | |
103 | */ | |
104 | /* | |
105 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 106 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
107 | * any extra contention... |
108 | */ | |
109 | ||
a02f76c3 | 110 | static int __link_path_walk(const char *name, struct nameidata *nd); |
c4a7808f | 111 | |
1da177e4 LT |
112 | /* In order to reduce some races, while at the same time doing additional |
113 | * checking and hopefully speeding things up, we copy filenames to the | |
114 | * kernel data space before using them.. | |
115 | * | |
116 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
117 | * PATH_MAX includes the nul terminator --RR. | |
118 | */ | |
858119e1 | 119 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
120 | { |
121 | int retval; | |
122 | unsigned long len = PATH_MAX; | |
123 | ||
124 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
125 | if ((unsigned long) filename >= TASK_SIZE) | |
126 | return -EFAULT; | |
127 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
128 | len = TASK_SIZE - (unsigned long) filename; | |
129 | } | |
130 | ||
131 | retval = strncpy_from_user(page, filename, len); | |
132 | if (retval > 0) { | |
133 | if (retval < len) | |
134 | return 0; | |
135 | return -ENAMETOOLONG; | |
136 | } else if (!retval) | |
137 | retval = -ENOENT; | |
138 | return retval; | |
139 | } | |
140 | ||
141 | char * getname(const char __user * filename) | |
142 | { | |
143 | char *tmp, *result; | |
144 | ||
145 | result = ERR_PTR(-ENOMEM); | |
146 | tmp = __getname(); | |
147 | if (tmp) { | |
148 | int retval = do_getname(filename, tmp); | |
149 | ||
150 | result = tmp; | |
151 | if (retval < 0) { | |
152 | __putname(tmp); | |
153 | result = ERR_PTR(retval); | |
154 | } | |
155 | } | |
156 | audit_getname(result); | |
157 | return result; | |
158 | } | |
159 | ||
160 | #ifdef CONFIG_AUDITSYSCALL | |
161 | void putname(const char *name) | |
162 | { | |
5ac3a9c2 | 163 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
164 | audit_putname(name); |
165 | else | |
166 | __putname(name); | |
167 | } | |
168 | EXPORT_SYMBOL(putname); | |
169 | #endif | |
170 | ||
171 | ||
172 | /** | |
173 | * generic_permission - check for access rights on a Posix-like filesystem | |
174 | * @inode: inode to check access rights for | |
175 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
176 | * @check_acl: optional callback to check for Posix ACLs | |
177 | * | |
178 | * Used to check for read/write/execute permissions on a file. | |
179 | * We use "fsuid" for this, letting us set arbitrary permissions | |
180 | * for filesystem access without changing the "normal" uids which | |
181 | * are used for other things.. | |
182 | */ | |
183 | int generic_permission(struct inode *inode, int mask, | |
184 | int (*check_acl)(struct inode *inode, int mask)) | |
185 | { | |
186 | umode_t mode = inode->i_mode; | |
187 | ||
e6305c43 AV |
188 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
189 | ||
da9592ed | 190 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
191 | mode >>= 6; |
192 | else { | |
193 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
194 | int error = check_acl(inode, mask); | |
195 | if (error == -EACCES) | |
196 | goto check_capabilities; | |
197 | else if (error != -EAGAIN) | |
198 | return error; | |
199 | } | |
200 | ||
201 | if (in_group_p(inode->i_gid)) | |
202 | mode >>= 3; | |
203 | } | |
204 | ||
205 | /* | |
206 | * If the DACs are ok we don't need any capability check. | |
207 | */ | |
e6305c43 | 208 | if ((mask & ~mode) == 0) |
1da177e4 LT |
209 | return 0; |
210 | ||
211 | check_capabilities: | |
212 | /* | |
213 | * Read/write DACs are always overridable. | |
214 | * Executable DACs are overridable if at least one exec bit is set. | |
215 | */ | |
f696a365 | 216 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
217 | if (capable(CAP_DAC_OVERRIDE)) |
218 | return 0; | |
219 | ||
220 | /* | |
221 | * Searching includes executable on directories, else just read. | |
222 | */ | |
223 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
224 | if (capable(CAP_DAC_READ_SEARCH)) | |
225 | return 0; | |
226 | ||
227 | return -EACCES; | |
228 | } | |
229 | ||
f419a2e3 | 230 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 231 | { |
e6305c43 | 232 | int retval; |
1da177e4 LT |
233 | |
234 | if (mask & MAY_WRITE) { | |
22590e41 | 235 | umode_t mode = inode->i_mode; |
1da177e4 LT |
236 | |
237 | /* | |
238 | * Nobody gets write access to a read-only fs. | |
239 | */ | |
240 | if (IS_RDONLY(inode) && | |
241 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
242 | return -EROFS; | |
243 | ||
244 | /* | |
245 | * Nobody gets write access to an immutable file. | |
246 | */ | |
247 | if (IS_IMMUTABLE(inode)) | |
248 | return -EACCES; | |
249 | } | |
250 | ||
1da177e4 | 251 | /* Ordinary permission routines do not understand MAY_APPEND. */ |
f696a365 | 252 | if (inode->i_op && inode->i_op->permission) |
b77b0646 | 253 | retval = inode->i_op->permission(inode, mask); |
f696a365 | 254 | else |
e6305c43 | 255 | retval = generic_permission(inode, mask, NULL); |
f696a365 | 256 | |
1da177e4 LT |
257 | if (retval) |
258 | return retval; | |
259 | ||
08ce5f16 SH |
260 | retval = devcgroup_inode_permission(inode, mask); |
261 | if (retval) | |
262 | return retval; | |
263 | ||
e6305c43 | 264 | return security_inode_permission(inode, |
f418b006 | 265 | mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND)); |
1da177e4 LT |
266 | } |
267 | ||
e4543edd CH |
268 | /** |
269 | * vfs_permission - check for access rights to a given path | |
270 | * @nd: lookup result that describes the path | |
271 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
272 | * | |
273 | * Used to check for read/write/execute permissions on a path. | |
274 | * We use "fsuid" for this, letting us set arbitrary permissions | |
275 | * for filesystem access without changing the "normal" uids which | |
276 | * are used for other things. | |
277 | */ | |
278 | int vfs_permission(struct nameidata *nd, int mask) | |
279 | { | |
f419a2e3 | 280 | return inode_permission(nd->path.dentry->d_inode, mask); |
e4543edd CH |
281 | } |
282 | ||
8c744fb8 CH |
283 | /** |
284 | * file_permission - check for additional access rights to a given file | |
285 | * @file: file to check access rights for | |
286 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
287 | * | |
288 | * Used to check for read/write/execute permissions on an already opened | |
289 | * file. | |
290 | * | |
291 | * Note: | |
292 | * Do not use this function in new code. All access checks should | |
293 | * be done using vfs_permission(). | |
294 | */ | |
295 | int file_permission(struct file *file, int mask) | |
296 | { | |
f419a2e3 | 297 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
298 | } |
299 | ||
1da177e4 LT |
300 | /* |
301 | * get_write_access() gets write permission for a file. | |
302 | * put_write_access() releases this write permission. | |
303 | * This is used for regular files. | |
304 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
305 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
306 | * can have the following values: | |
307 | * 0: no writers, no VM_DENYWRITE mappings | |
308 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
309 | * > 0: (i_writecount) users are writing to the file. | |
310 | * | |
311 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
312 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
313 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
314 | * to do the change if sign is wrong. Exclusion between them is provided by | |
315 | * the inode->i_lock spinlock. | |
316 | */ | |
317 | ||
318 | int get_write_access(struct inode * inode) | |
319 | { | |
320 | spin_lock(&inode->i_lock); | |
321 | if (atomic_read(&inode->i_writecount) < 0) { | |
322 | spin_unlock(&inode->i_lock); | |
323 | return -ETXTBSY; | |
324 | } | |
325 | atomic_inc(&inode->i_writecount); | |
326 | spin_unlock(&inode->i_lock); | |
327 | ||
328 | return 0; | |
329 | } | |
330 | ||
331 | int deny_write_access(struct file * file) | |
332 | { | |
0f7fc9e4 | 333 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
334 | |
335 | spin_lock(&inode->i_lock); | |
336 | if (atomic_read(&inode->i_writecount) > 0) { | |
337 | spin_unlock(&inode->i_lock); | |
338 | return -ETXTBSY; | |
339 | } | |
340 | atomic_dec(&inode->i_writecount); | |
341 | spin_unlock(&inode->i_lock); | |
342 | ||
343 | return 0; | |
344 | } | |
345 | ||
5dd784d0 JB |
346 | /** |
347 | * path_get - get a reference to a path | |
348 | * @path: path to get the reference to | |
349 | * | |
350 | * Given a path increment the reference count to the dentry and the vfsmount. | |
351 | */ | |
352 | void path_get(struct path *path) | |
353 | { | |
354 | mntget(path->mnt); | |
355 | dget(path->dentry); | |
356 | } | |
357 | EXPORT_SYMBOL(path_get); | |
358 | ||
1d957f9b JB |
359 | /** |
360 | * path_put - put a reference to a path | |
361 | * @path: path to put the reference to | |
362 | * | |
363 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
364 | */ | |
365 | void path_put(struct path *path) | |
1da177e4 | 366 | { |
1d957f9b JB |
367 | dput(path->dentry); |
368 | mntput(path->mnt); | |
1da177e4 | 369 | } |
1d957f9b | 370 | EXPORT_SYMBOL(path_put); |
1da177e4 | 371 | |
834f2a4a TM |
372 | /** |
373 | * release_open_intent - free up open intent resources | |
374 | * @nd: pointer to nameidata | |
375 | */ | |
376 | void release_open_intent(struct nameidata *nd) | |
377 | { | |
0f7fc9e4 | 378 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
379 | put_filp(nd->intent.open.file); |
380 | else | |
381 | fput(nd->intent.open.file); | |
382 | } | |
383 | ||
bcdc5e01 IK |
384 | static inline struct dentry * |
385 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
386 | { | |
387 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
388 | if (unlikely(status <= 0)) { | |
389 | /* | |
390 | * The dentry failed validation. | |
391 | * If d_revalidate returned 0 attempt to invalidate | |
392 | * the dentry otherwise d_revalidate is asking us | |
393 | * to return a fail status. | |
394 | */ | |
395 | if (!status) { | |
396 | if (!d_invalidate(dentry)) { | |
397 | dput(dentry); | |
398 | dentry = NULL; | |
399 | } | |
400 | } else { | |
401 | dput(dentry); | |
402 | dentry = ERR_PTR(status); | |
403 | } | |
404 | } | |
405 | return dentry; | |
406 | } | |
407 | ||
1da177e4 LT |
408 | /* |
409 | * Internal lookup() using the new generic dcache. | |
410 | * SMP-safe | |
411 | */ | |
412 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
413 | { | |
414 | struct dentry * dentry = __d_lookup(parent, name); | |
415 | ||
416 | /* lockess __d_lookup may fail due to concurrent d_move() | |
417 | * in some unrelated directory, so try with d_lookup | |
418 | */ | |
419 | if (!dentry) | |
420 | dentry = d_lookup(parent, name); | |
421 | ||
bcdc5e01 IK |
422 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
423 | dentry = do_revalidate(dentry, nd); | |
424 | ||
1da177e4 LT |
425 | return dentry; |
426 | } | |
427 | ||
428 | /* | |
429 | * Short-cut version of permission(), for calling by | |
430 | * path_walk(), when dcache lock is held. Combines parts | |
431 | * of permission() and generic_permission(), and tests ONLY for | |
432 | * MAY_EXEC permission. | |
433 | * | |
434 | * If appropriate, check DAC only. If not appropriate, or | |
435 | * short-cut DAC fails, then call permission() to do more | |
436 | * complete permission check. | |
437 | */ | |
672b16b2 | 438 | static int exec_permission_lite(struct inode *inode) |
1da177e4 LT |
439 | { |
440 | umode_t mode = inode->i_mode; | |
441 | ||
442 | if (inode->i_op && inode->i_op->permission) | |
443 | return -EAGAIN; | |
444 | ||
da9592ed | 445 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
446 | mode >>= 6; |
447 | else if (in_group_p(inode->i_gid)) | |
448 | mode >>= 3; | |
449 | ||
450 | if (mode & MAY_EXEC) | |
451 | goto ok; | |
452 | ||
453 | if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE)) | |
454 | goto ok; | |
455 | ||
456 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE)) | |
457 | goto ok; | |
458 | ||
459 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH)) | |
460 | goto ok; | |
461 | ||
462 | return -EACCES; | |
463 | ok: | |
b77b0646 | 464 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
465 | } |
466 | ||
467 | /* | |
468 | * This is called when everything else fails, and we actually have | |
469 | * to go to the low-level filesystem to find out what we should do.. | |
470 | * | |
471 | * We get the directory semaphore, and after getting that we also | |
472 | * make sure that nobody added the entry to the dcache in the meantime.. | |
473 | * SMP-safe | |
474 | */ | |
475 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
476 | { | |
477 | struct dentry * result; | |
478 | struct inode *dir = parent->d_inode; | |
479 | ||
1b1dcc1b | 480 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
481 | /* |
482 | * First re-do the cached lookup just in case it was created | |
483 | * while we waited for the directory semaphore.. | |
484 | * | |
485 | * FIXME! This could use version numbering or similar to | |
486 | * avoid unnecessary cache lookups. | |
487 | * | |
488 | * The "dcache_lock" is purely to protect the RCU list walker | |
489 | * from concurrent renames at this point (we mustn't get false | |
490 | * negatives from the RCU list walk here, unlike the optimistic | |
491 | * fast walk). | |
492 | * | |
493 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
494 | */ | |
495 | result = d_lookup(parent, name); | |
496 | if (!result) { | |
d70b67c8 MS |
497 | struct dentry *dentry; |
498 | ||
499 | /* Don't create child dentry for a dead directory. */ | |
500 | result = ERR_PTR(-ENOENT); | |
501 | if (IS_DEADDIR(dir)) | |
502 | goto out_unlock; | |
503 | ||
504 | dentry = d_alloc(parent, name); | |
1da177e4 LT |
505 | result = ERR_PTR(-ENOMEM); |
506 | if (dentry) { | |
507 | result = dir->i_op->lookup(dir, dentry, nd); | |
508 | if (result) | |
509 | dput(dentry); | |
510 | else | |
511 | result = dentry; | |
512 | } | |
d70b67c8 | 513 | out_unlock: |
1b1dcc1b | 514 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
515 | return result; |
516 | } | |
517 | ||
518 | /* | |
519 | * Uhhuh! Nasty case: the cache was re-populated while | |
520 | * we waited on the semaphore. Need to revalidate. | |
521 | */ | |
1b1dcc1b | 522 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 523 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
524 | result = do_revalidate(result, nd); |
525 | if (!result) | |
1da177e4 | 526 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
527 | } |
528 | return result; | |
529 | } | |
530 | ||
1da177e4 | 531 | /* SMP-safe */ |
7f2da1e7 | 532 | static __always_inline void |
1da177e4 LT |
533 | walk_init_root(const char *name, struct nameidata *nd) |
534 | { | |
e518ddb7 AM |
535 | struct fs_struct *fs = current->fs; |
536 | ||
537 | read_lock(&fs->lock); | |
6ac08c39 JB |
538 | nd->path = fs->root; |
539 | path_get(&fs->root); | |
e518ddb7 | 540 | read_unlock(&fs->lock); |
1da177e4 LT |
541 | } |
542 | ||
a02f76c3 AV |
543 | /* |
544 | * Wrapper to retry pathname resolution whenever the underlying | |
545 | * file system returns an ESTALE. | |
546 | * | |
547 | * Retry the whole path once, forcing real lookup requests | |
548 | * instead of relying on the dcache. | |
549 | */ | |
550 | static __always_inline int link_path_walk(const char *name, struct nameidata *nd) | |
551 | { | |
552 | struct path save = nd->path; | |
553 | int result; | |
554 | ||
555 | /* make sure the stuff we saved doesn't go away */ | |
c8e7f449 | 556 | path_get(&save); |
a02f76c3 AV |
557 | |
558 | result = __link_path_walk(name, nd); | |
559 | if (result == -ESTALE) { | |
560 | /* nd->path had been dropped */ | |
561 | nd->path = save; | |
c8e7f449 | 562 | path_get(&nd->path); |
a02f76c3 AV |
563 | nd->flags |= LOOKUP_REVAL; |
564 | result = __link_path_walk(name, nd); | |
565 | } | |
566 | ||
567 | path_put(&save); | |
568 | ||
569 | return result; | |
570 | } | |
571 | ||
f1662356 | 572 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
573 | { |
574 | int res = 0; | |
575 | char *name; | |
576 | if (IS_ERR(link)) | |
577 | goto fail; | |
578 | ||
579 | if (*link == '/') { | |
1d957f9b | 580 | path_put(&nd->path); |
7f2da1e7 | 581 | walk_init_root(link, nd); |
1da177e4 LT |
582 | } |
583 | res = link_path_walk(link, nd); | |
1da177e4 LT |
584 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
585 | return res; | |
586 | /* | |
587 | * If it is an iterative symlinks resolution in open_namei() we | |
588 | * have to copy the last component. And all that crap because of | |
589 | * bloody create() on broken symlinks. Furrfu... | |
590 | */ | |
591 | name = __getname(); | |
592 | if (unlikely(!name)) { | |
1d957f9b | 593 | path_put(&nd->path); |
1da177e4 LT |
594 | return -ENOMEM; |
595 | } | |
596 | strcpy(name, nd->last.name); | |
597 | nd->last.name = name; | |
598 | return 0; | |
599 | fail: | |
1d957f9b | 600 | path_put(&nd->path); |
1da177e4 LT |
601 | return PTR_ERR(link); |
602 | } | |
603 | ||
1d957f9b | 604 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
605 | { |
606 | dput(path->dentry); | |
4ac91378 | 607 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
608 | mntput(path->mnt); |
609 | } | |
610 | ||
611 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
612 | { | |
4ac91378 JB |
613 | dput(nd->path.dentry); |
614 | if (nd->path.mnt != path->mnt) | |
615 | mntput(nd->path.mnt); | |
616 | nd->path.mnt = path->mnt; | |
617 | nd->path.dentry = path->dentry; | |
051d3812 IK |
618 | } |
619 | ||
f1662356 | 620 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
621 | { |
622 | int error; | |
cc314eef | 623 | void *cookie; |
cd4e91d3 | 624 | struct dentry *dentry = path->dentry; |
1da177e4 | 625 | |
d671a1cb | 626 | touch_atime(path->mnt, dentry); |
1da177e4 | 627 | nd_set_link(nd, NULL); |
cd4e91d3 | 628 | |
4ac91378 | 629 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
630 | path_to_nameidata(path, nd); |
631 | dget(dentry); | |
632 | } | |
633 | mntget(path->mnt); | |
cc314eef LT |
634 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
635 | error = PTR_ERR(cookie); | |
636 | if (!IS_ERR(cookie)) { | |
1da177e4 | 637 | char *s = nd_get_link(nd); |
cc314eef | 638 | error = 0; |
1da177e4 LT |
639 | if (s) |
640 | error = __vfs_follow_link(nd, s); | |
641 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 642 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 643 | } |
09da5916 | 644 | path_put(path); |
1da177e4 LT |
645 | |
646 | return error; | |
647 | } | |
648 | ||
649 | /* | |
650 | * This limits recursive symlink follows to 8, while | |
651 | * limiting consecutive symlinks to 40. | |
652 | * | |
653 | * Without that kind of total limit, nasty chains of consecutive | |
654 | * symlinks can cause almost arbitrarily long lookups. | |
655 | */ | |
90ebe565 | 656 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
657 | { |
658 | int err = -ELOOP; | |
659 | if (current->link_count >= MAX_NESTED_LINKS) | |
660 | goto loop; | |
661 | if (current->total_link_count >= 40) | |
662 | goto loop; | |
663 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
664 | cond_resched(); | |
90ebe565 | 665 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
666 | if (err) |
667 | goto loop; | |
668 | current->link_count++; | |
669 | current->total_link_count++; | |
670 | nd->depth++; | |
cd4e91d3 | 671 | err = __do_follow_link(path, nd); |
839d9f93 AV |
672 | current->link_count--; |
673 | nd->depth--; | |
1da177e4 LT |
674 | return err; |
675 | loop: | |
1d957f9b JB |
676 | path_put_conditional(path, nd); |
677 | path_put(&nd->path); | |
1da177e4 LT |
678 | return err; |
679 | } | |
680 | ||
681 | int follow_up(struct vfsmount **mnt, struct dentry **dentry) | |
682 | { | |
683 | struct vfsmount *parent; | |
684 | struct dentry *mountpoint; | |
685 | spin_lock(&vfsmount_lock); | |
686 | parent=(*mnt)->mnt_parent; | |
687 | if (parent == *mnt) { | |
688 | spin_unlock(&vfsmount_lock); | |
689 | return 0; | |
690 | } | |
691 | mntget(parent); | |
692 | mountpoint=dget((*mnt)->mnt_mountpoint); | |
693 | spin_unlock(&vfsmount_lock); | |
694 | dput(*dentry); | |
695 | *dentry = mountpoint; | |
696 | mntput(*mnt); | |
697 | *mnt = parent; | |
698 | return 1; | |
699 | } | |
700 | ||
701 | /* no need for dcache_lock, as serialization is taken care in | |
702 | * namespace.c | |
703 | */ | |
463ffb2e AV |
704 | static int __follow_mount(struct path *path) |
705 | { | |
706 | int res = 0; | |
707 | while (d_mountpoint(path->dentry)) { | |
708 | struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry); | |
709 | if (!mounted) | |
710 | break; | |
711 | dput(path->dentry); | |
712 | if (res) | |
713 | mntput(path->mnt); | |
714 | path->mnt = mounted; | |
715 | path->dentry = dget(mounted->mnt_root); | |
716 | res = 1; | |
717 | } | |
718 | return res; | |
719 | } | |
720 | ||
58c465eb | 721 | static void follow_mount(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 | 722 | { |
1da177e4 LT |
723 | while (d_mountpoint(*dentry)) { |
724 | struct vfsmount *mounted = lookup_mnt(*mnt, *dentry); | |
725 | if (!mounted) | |
726 | break; | |
58c465eb | 727 | dput(*dentry); |
1da177e4 LT |
728 | mntput(*mnt); |
729 | *mnt = mounted; | |
1da177e4 | 730 | *dentry = dget(mounted->mnt_root); |
1da177e4 | 731 | } |
1da177e4 LT |
732 | } |
733 | ||
734 | /* no need for dcache_lock, as serialization is taken care in | |
735 | * namespace.c | |
736 | */ | |
e13b210f | 737 | int follow_down(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 LT |
738 | { |
739 | struct vfsmount *mounted; | |
740 | ||
741 | mounted = lookup_mnt(*mnt, *dentry); | |
742 | if (mounted) { | |
e13b210f | 743 | dput(*dentry); |
1da177e4 LT |
744 | mntput(*mnt); |
745 | *mnt = mounted; | |
1da177e4 LT |
746 | *dentry = dget(mounted->mnt_root); |
747 | return 1; | |
748 | } | |
749 | return 0; | |
750 | } | |
751 | ||
f1662356 | 752 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 753 | { |
e518ddb7 AM |
754 | struct fs_struct *fs = current->fs; |
755 | ||
1da177e4 LT |
756 | while(1) { |
757 | struct vfsmount *parent; | |
4ac91378 | 758 | struct dentry *old = nd->path.dentry; |
1da177e4 | 759 | |
e518ddb7 | 760 | read_lock(&fs->lock); |
6ac08c39 JB |
761 | if (nd->path.dentry == fs->root.dentry && |
762 | nd->path.mnt == fs->root.mnt) { | |
e518ddb7 | 763 | read_unlock(&fs->lock); |
1da177e4 LT |
764 | break; |
765 | } | |
e518ddb7 | 766 | read_unlock(&fs->lock); |
1da177e4 | 767 | spin_lock(&dcache_lock); |
4ac91378 JB |
768 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
769 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
770 | spin_unlock(&dcache_lock); |
771 | dput(old); | |
772 | break; | |
773 | } | |
774 | spin_unlock(&dcache_lock); | |
775 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
776 | parent = nd->path.mnt->mnt_parent; |
777 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
778 | spin_unlock(&vfsmount_lock); |
779 | break; | |
780 | } | |
781 | mntget(parent); | |
4ac91378 | 782 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
783 | spin_unlock(&vfsmount_lock); |
784 | dput(old); | |
4ac91378 JB |
785 | mntput(nd->path.mnt); |
786 | nd->path.mnt = parent; | |
1da177e4 | 787 | } |
4ac91378 | 788 | follow_mount(&nd->path.mnt, &nd->path.dentry); |
1da177e4 LT |
789 | } |
790 | ||
1da177e4 LT |
791 | /* |
792 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
793 | * small and for now I'd prefer to have fast path as straight as possible. | |
794 | * It _is_ time-critical. | |
795 | */ | |
796 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
797 | struct path *path) | |
798 | { | |
4ac91378 JB |
799 | struct vfsmount *mnt = nd->path.mnt; |
800 | struct dentry *dentry = __d_lookup(nd->path.dentry, name); | |
1da177e4 LT |
801 | |
802 | if (!dentry) | |
803 | goto need_lookup; | |
804 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
805 | goto need_revalidate; | |
806 | done: | |
807 | path->mnt = mnt; | |
808 | path->dentry = dentry; | |
634ee701 | 809 | __follow_mount(path); |
1da177e4 LT |
810 | return 0; |
811 | ||
812 | need_lookup: | |
4ac91378 | 813 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
814 | if (IS_ERR(dentry)) |
815 | goto fail; | |
816 | goto done; | |
817 | ||
818 | need_revalidate: | |
bcdc5e01 IK |
819 | dentry = do_revalidate(dentry, nd); |
820 | if (!dentry) | |
821 | goto need_lookup; | |
822 | if (IS_ERR(dentry)) | |
823 | goto fail; | |
824 | goto done; | |
1da177e4 LT |
825 | |
826 | fail: | |
827 | return PTR_ERR(dentry); | |
828 | } | |
829 | ||
830 | /* | |
831 | * Name resolution. | |
ea3834d9 PM |
832 | * This is the basic name resolution function, turning a pathname into |
833 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 834 | * |
ea3834d9 PM |
835 | * Returns 0 and nd will have valid dentry and mnt on success. |
836 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 837 | */ |
fc9b52cd | 838 | static int __link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
839 | { |
840 | struct path next; | |
841 | struct inode *inode; | |
842 | int err; | |
843 | unsigned int lookup_flags = nd->flags; | |
844 | ||
845 | while (*name=='/') | |
846 | name++; | |
847 | if (!*name) | |
848 | goto return_reval; | |
849 | ||
4ac91378 | 850 | inode = nd->path.dentry->d_inode; |
1da177e4 | 851 | if (nd->depth) |
f55eab82 | 852 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
853 | |
854 | /* At this point we know we have a real path component. */ | |
855 | for(;;) { | |
856 | unsigned long hash; | |
857 | struct qstr this; | |
858 | unsigned int c; | |
859 | ||
cdce5d6b | 860 | nd->flags |= LOOKUP_CONTINUE; |
672b16b2 | 861 | err = exec_permission_lite(inode); |
e4543edd CH |
862 | if (err == -EAGAIN) |
863 | err = vfs_permission(nd, MAY_EXEC); | |
6146f0d5 MZ |
864 | if (!err) |
865 | err = ima_path_check(&nd->path, MAY_EXEC); | |
1da177e4 LT |
866 | if (err) |
867 | break; | |
868 | ||
869 | this.name = name; | |
870 | c = *(const unsigned char *)name; | |
871 | ||
872 | hash = init_name_hash(); | |
873 | do { | |
874 | name++; | |
875 | hash = partial_name_hash(c, hash); | |
876 | c = *(const unsigned char *)name; | |
877 | } while (c && (c != '/')); | |
878 | this.len = name - (const char *) this.name; | |
879 | this.hash = end_name_hash(hash); | |
880 | ||
881 | /* remove trailing slashes? */ | |
882 | if (!c) | |
883 | goto last_component; | |
884 | while (*++name == '/'); | |
885 | if (!*name) | |
886 | goto last_with_slashes; | |
887 | ||
888 | /* | |
889 | * "." and ".." are special - ".." especially so because it has | |
890 | * to be able to know about the current root directory and | |
891 | * parent relationships. | |
892 | */ | |
893 | if (this.name[0] == '.') switch (this.len) { | |
894 | default: | |
895 | break; | |
896 | case 2: | |
897 | if (this.name[1] != '.') | |
898 | break; | |
58c465eb | 899 | follow_dotdot(nd); |
4ac91378 | 900 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
901 | /* fallthrough */ |
902 | case 1: | |
903 | continue; | |
904 | } | |
905 | /* | |
906 | * See if the low-level filesystem might want | |
907 | * to use its own hash.. | |
908 | */ | |
4ac91378 JB |
909 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
910 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
911 | &this); | |
1da177e4 LT |
912 | if (err < 0) |
913 | break; | |
914 | } | |
1da177e4 LT |
915 | /* This does the actual lookups.. */ |
916 | err = do_lookup(nd, &this, &next); | |
917 | if (err) | |
918 | break; | |
1da177e4 LT |
919 | |
920 | err = -ENOENT; | |
921 | inode = next.dentry->d_inode; | |
922 | if (!inode) | |
923 | goto out_dput; | |
924 | err = -ENOTDIR; | |
925 | if (!inode->i_op) | |
926 | goto out_dput; | |
927 | ||
928 | if (inode->i_op->follow_link) { | |
90ebe565 | 929 | err = do_follow_link(&next, nd); |
1da177e4 LT |
930 | if (err) |
931 | goto return_err; | |
932 | err = -ENOENT; | |
4ac91378 | 933 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
934 | if (!inode) |
935 | break; | |
936 | err = -ENOTDIR; | |
937 | if (!inode->i_op) | |
938 | break; | |
09dd17d3 MS |
939 | } else |
940 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
941 | err = -ENOTDIR; |
942 | if (!inode->i_op->lookup) | |
943 | break; | |
944 | continue; | |
945 | /* here ends the main loop */ | |
946 | ||
947 | last_with_slashes: | |
948 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
949 | last_component: | |
f55eab82 TM |
950 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
951 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
952 | if (lookup_flags & LOOKUP_PARENT) |
953 | goto lookup_parent; | |
954 | if (this.name[0] == '.') switch (this.len) { | |
955 | default: | |
956 | break; | |
957 | case 2: | |
958 | if (this.name[1] != '.') | |
959 | break; | |
58c465eb | 960 | follow_dotdot(nd); |
4ac91378 | 961 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
962 | /* fallthrough */ |
963 | case 1: | |
964 | goto return_reval; | |
965 | } | |
4ac91378 JB |
966 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
967 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
968 | &this); | |
1da177e4 LT |
969 | if (err < 0) |
970 | break; | |
971 | } | |
972 | err = do_lookup(nd, &this, &next); | |
973 | if (err) | |
974 | break; | |
1da177e4 LT |
975 | inode = next.dentry->d_inode; |
976 | if ((lookup_flags & LOOKUP_FOLLOW) | |
977 | && inode && inode->i_op && inode->i_op->follow_link) { | |
90ebe565 | 978 | err = do_follow_link(&next, nd); |
1da177e4 LT |
979 | if (err) |
980 | goto return_err; | |
4ac91378 | 981 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
982 | } else |
983 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
984 | err = -ENOENT; |
985 | if (!inode) | |
986 | break; | |
987 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
988 | err = -ENOTDIR; | |
989 | if (!inode->i_op || !inode->i_op->lookup) | |
990 | break; | |
991 | } | |
992 | goto return_base; | |
993 | lookup_parent: | |
994 | nd->last = this; | |
995 | nd->last_type = LAST_NORM; | |
996 | if (this.name[0] != '.') | |
997 | goto return_base; | |
998 | if (this.len == 1) | |
999 | nd->last_type = LAST_DOT; | |
1000 | else if (this.len == 2 && this.name[1] == '.') | |
1001 | nd->last_type = LAST_DOTDOT; | |
1002 | else | |
1003 | goto return_base; | |
1004 | return_reval: | |
1005 | /* | |
1006 | * We bypassed the ordinary revalidation routines. | |
1007 | * We may need to check the cached dentry for staleness. | |
1008 | */ | |
4ac91378 JB |
1009 | if (nd->path.dentry && nd->path.dentry->d_sb && |
1010 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
1011 | err = -ESTALE; |
1012 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
1013 | if (!nd->path.dentry->d_op->d_revalidate( |
1014 | nd->path.dentry, nd)) | |
1da177e4 LT |
1015 | break; |
1016 | } | |
1017 | return_base: | |
1018 | return 0; | |
1019 | out_dput: | |
1d957f9b | 1020 | path_put_conditional(&next, nd); |
1da177e4 LT |
1021 | break; |
1022 | } | |
1d957f9b | 1023 | path_put(&nd->path); |
1da177e4 LT |
1024 | return_err: |
1025 | return err; | |
1026 | } | |
1027 | ||
fc9b52cd | 1028 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1029 | { |
1030 | current->total_link_count = 0; | |
1031 | return link_path_walk(name, nd); | |
1032 | } | |
1033 | ||
ea3834d9 | 1034 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
fc9b52cd | 1035 | static int do_path_lookup(int dfd, const char *name, |
5590ff0d | 1036 | unsigned int flags, struct nameidata *nd) |
1da177e4 | 1037 | { |
ea3834d9 | 1038 | int retval = 0; |
170aa3d0 UD |
1039 | int fput_needed; |
1040 | struct file *file; | |
e518ddb7 | 1041 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1042 | |
1043 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1044 | nd->flags = flags; | |
1045 | nd->depth = 0; | |
1046 | ||
1da177e4 | 1047 | if (*name=='/') { |
e518ddb7 | 1048 | read_lock(&fs->lock); |
6ac08c39 JB |
1049 | nd->path = fs->root; |
1050 | path_get(&fs->root); | |
e518ddb7 | 1051 | read_unlock(&fs->lock); |
5590ff0d | 1052 | } else if (dfd == AT_FDCWD) { |
e518ddb7 | 1053 | read_lock(&fs->lock); |
6ac08c39 JB |
1054 | nd->path = fs->pwd; |
1055 | path_get(&fs->pwd); | |
e518ddb7 | 1056 | read_unlock(&fs->lock); |
5590ff0d | 1057 | } else { |
5590ff0d UD |
1058 | struct dentry *dentry; |
1059 | ||
1060 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1061 | retval = -EBADF; |
1062 | if (!file) | |
6d09bb62 | 1063 | goto out_fail; |
5590ff0d | 1064 | |
0f7fc9e4 | 1065 | dentry = file->f_path.dentry; |
5590ff0d | 1066 | |
170aa3d0 UD |
1067 | retval = -ENOTDIR; |
1068 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1069 | goto fput_fail; |
5590ff0d UD |
1070 | |
1071 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1072 | if (retval) |
6d09bb62 | 1073 | goto fput_fail; |
5590ff0d | 1074 | |
5dd784d0 JB |
1075 | nd->path = file->f_path; |
1076 | path_get(&file->f_path); | |
5590ff0d UD |
1077 | |
1078 | fput_light(file, fput_needed); | |
1da177e4 | 1079 | } |
2dfdd266 JJS |
1080 | |
1081 | retval = path_walk(name, nd); | |
4ac91378 JB |
1082 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1083 | nd->path.dentry->d_inode)) | |
1084 | audit_inode(name, nd->path.dentry); | |
6d09bb62 | 1085 | out_fail: |
170aa3d0 UD |
1086 | return retval; |
1087 | ||
6d09bb62 | 1088 | fput_fail: |
170aa3d0 | 1089 | fput_light(file, fput_needed); |
6d09bb62 | 1090 | goto out_fail; |
1da177e4 LT |
1091 | } |
1092 | ||
fc9b52cd | 1093 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1094 | struct nameidata *nd) |
1095 | { | |
1096 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1097 | } | |
1098 | ||
d1811465 AV |
1099 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1100 | { | |
1101 | struct nameidata nd; | |
1102 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1103 | if (!res) | |
1104 | *path = nd.path; | |
1105 | return res; | |
1106 | } | |
1107 | ||
16f18200 JJS |
1108 | /** |
1109 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1110 | * @dentry: pointer to dentry of the base directory | |
1111 | * @mnt: pointer to vfs mount of the base directory | |
1112 | * @name: pointer to file name | |
1113 | * @flags: lookup flags | |
1114 | * @nd: pointer to nameidata | |
1115 | */ | |
1116 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1117 | const char *name, unsigned int flags, | |
1118 | struct nameidata *nd) | |
1119 | { | |
1120 | int retval; | |
1121 | ||
1122 | /* same as do_path_lookup */ | |
1123 | nd->last_type = LAST_ROOT; | |
1124 | nd->flags = flags; | |
1125 | nd->depth = 0; | |
1126 | ||
c8e7f449 JB |
1127 | nd->path.dentry = dentry; |
1128 | nd->path.mnt = mnt; | |
1129 | path_get(&nd->path); | |
16f18200 JJS |
1130 | |
1131 | retval = path_walk(name, nd); | |
4ac91378 JB |
1132 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1133 | nd->path.dentry->d_inode)) | |
1134 | audit_inode(name, nd->path.dentry); | |
16f18200 JJS |
1135 | |
1136 | return retval; | |
1137 | ||
1138 | } | |
1139 | ||
8737f3a1 AV |
1140 | /** |
1141 | * path_lookup_open - lookup a file path with open intent | |
1142 | * @dfd: the directory to use as base, or AT_FDCWD | |
1143 | * @name: pointer to file name | |
1144 | * @lookup_flags: lookup intent flags | |
1145 | * @nd: pointer to nameidata | |
1146 | * @open_flags: open intent flags | |
1147 | */ | |
1148 | int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags, | |
1149 | struct nameidata *nd, int open_flags) | |
834f2a4a TM |
1150 | { |
1151 | struct file *filp = get_empty_filp(); | |
1152 | int err; | |
1153 | ||
1154 | if (filp == NULL) | |
1155 | return -ENFILE; | |
1156 | nd->intent.open.file = filp; | |
1157 | nd->intent.open.flags = open_flags; | |
8737f3a1 | 1158 | nd->intent.open.create_mode = 0; |
5590ff0d | 1159 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1160 | if (IS_ERR(nd->intent.open.file)) { |
1161 | if (err == 0) { | |
1162 | err = PTR_ERR(nd->intent.open.file); | |
1d957f9b | 1163 | path_put(&nd->path); |
834f2a4a TM |
1164 | } |
1165 | } else if (err != 0) | |
1166 | release_open_intent(nd); | |
1167 | return err; | |
1168 | } | |
1169 | ||
eead1911 CH |
1170 | static struct dentry *__lookup_hash(struct qstr *name, |
1171 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1172 | { |
057f6c01 | 1173 | struct dentry *dentry; |
1da177e4 LT |
1174 | struct inode *inode; |
1175 | int err; | |
1176 | ||
1177 | inode = base->d_inode; | |
1da177e4 LT |
1178 | |
1179 | /* | |
1180 | * See if the low-level filesystem might want | |
1181 | * to use its own hash.. | |
1182 | */ | |
1183 | if (base->d_op && base->d_op->d_hash) { | |
1184 | err = base->d_op->d_hash(base, name); | |
1185 | dentry = ERR_PTR(err); | |
1186 | if (err < 0) | |
1187 | goto out; | |
1188 | } | |
1189 | ||
1190 | dentry = cached_lookup(base, name, nd); | |
1191 | if (!dentry) { | |
d70b67c8 MS |
1192 | struct dentry *new; |
1193 | ||
1194 | /* Don't create child dentry for a dead directory. */ | |
1195 | dentry = ERR_PTR(-ENOENT); | |
1196 | if (IS_DEADDIR(inode)) | |
1197 | goto out; | |
1198 | ||
1199 | new = d_alloc(base, name); | |
1da177e4 LT |
1200 | dentry = ERR_PTR(-ENOMEM); |
1201 | if (!new) | |
1202 | goto out; | |
1203 | dentry = inode->i_op->lookup(inode, new, nd); | |
1204 | if (!dentry) | |
1205 | dentry = new; | |
1206 | else | |
1207 | dput(new); | |
1208 | } | |
1209 | out: | |
1210 | return dentry; | |
1211 | } | |
1212 | ||
057f6c01 JM |
1213 | /* |
1214 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1215 | * needs parent already locked. Doesn't follow mounts. | |
1216 | * SMP-safe. | |
1217 | */ | |
eead1911 | 1218 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1219 | { |
057f6c01 JM |
1220 | int err; |
1221 | ||
f419a2e3 | 1222 | err = inode_permission(nd->path.dentry->d_inode, MAY_EXEC); |
057f6c01 | 1223 | if (err) |
eead1911 | 1224 | return ERR_PTR(err); |
4ac91378 | 1225 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1226 | } |
1227 | ||
eead1911 CH |
1228 | static int __lookup_one_len(const char *name, struct qstr *this, |
1229 | struct dentry *base, int len) | |
1da177e4 LT |
1230 | { |
1231 | unsigned long hash; | |
1da177e4 LT |
1232 | unsigned int c; |
1233 | ||
057f6c01 JM |
1234 | this->name = name; |
1235 | this->len = len; | |
1da177e4 | 1236 | if (!len) |
057f6c01 | 1237 | return -EACCES; |
1da177e4 LT |
1238 | |
1239 | hash = init_name_hash(); | |
1240 | while (len--) { | |
1241 | c = *(const unsigned char *)name++; | |
1242 | if (c == '/' || c == '\0') | |
057f6c01 | 1243 | return -EACCES; |
1da177e4 LT |
1244 | hash = partial_name_hash(c, hash); |
1245 | } | |
057f6c01 JM |
1246 | this->hash = end_name_hash(hash); |
1247 | return 0; | |
1248 | } | |
1da177e4 | 1249 | |
eead1911 | 1250 | /** |
a6b91919 | 1251 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1252 | * @name: pathname component to lookup |
1253 | * @base: base directory to lookup from | |
1254 | * @len: maximum length @len should be interpreted to | |
1255 | * | |
a6b91919 RD |
1256 | * Note that this routine is purely a helper for filesystem usage and should |
1257 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1258 | * nameidata argument is passed to the filesystem methods and a filesystem |
1259 | * using this helper needs to be prepared for that. | |
1260 | */ | |
057f6c01 JM |
1261 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1262 | { | |
1263 | int err; | |
1264 | struct qstr this; | |
1265 | ||
1266 | err = __lookup_one_len(name, &this, base, len); | |
eead1911 CH |
1267 | if (err) |
1268 | return ERR_PTR(err); | |
1269 | ||
f419a2e3 | 1270 | err = inode_permission(base->d_inode, MAY_EXEC); |
057f6c01 JM |
1271 | if (err) |
1272 | return ERR_PTR(err); | |
49705b77 | 1273 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1274 | } |
1275 | ||
eead1911 CH |
1276 | /** |
1277 | * lookup_one_noperm - bad hack for sysfs | |
1278 | * @name: pathname component to lookup | |
1279 | * @base: base directory to lookup from | |
1280 | * | |
1281 | * This is a variant of lookup_one_len that doesn't perform any permission | |
1282 | * checks. It's a horrible hack to work around the braindead sysfs | |
1283 | * architecture and should not be used anywhere else. | |
1284 | * | |
1285 | * DON'T USE THIS FUNCTION EVER, thanks. | |
1286 | */ | |
1287 | struct dentry *lookup_one_noperm(const char *name, struct dentry *base) | |
057f6c01 JM |
1288 | { |
1289 | int err; | |
1290 | struct qstr this; | |
1291 | ||
eead1911 | 1292 | err = __lookup_one_len(name, &this, base, strlen(name)); |
057f6c01 JM |
1293 | if (err) |
1294 | return ERR_PTR(err); | |
eead1911 | 1295 | return __lookup_hash(&this, base, NULL); |
1da177e4 LT |
1296 | } |
1297 | ||
2d8f3038 AV |
1298 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1299 | struct path *path) | |
1da177e4 | 1300 | { |
2d8f3038 | 1301 | struct nameidata nd; |
1da177e4 LT |
1302 | char *tmp = getname(name); |
1303 | int err = PTR_ERR(tmp); | |
1da177e4 | 1304 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1305 | |
1306 | BUG_ON(flags & LOOKUP_PARENT); | |
1307 | ||
1308 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1309 | putname(tmp); |
2d8f3038 AV |
1310 | if (!err) |
1311 | *path = nd.path; | |
1da177e4 LT |
1312 | } |
1313 | return err; | |
1314 | } | |
1315 | ||
2ad94ae6 AV |
1316 | static int user_path_parent(int dfd, const char __user *path, |
1317 | struct nameidata *nd, char **name) | |
1318 | { | |
1319 | char *s = getname(path); | |
1320 | int error; | |
1321 | ||
1322 | if (IS_ERR(s)) | |
1323 | return PTR_ERR(s); | |
1324 | ||
1325 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1326 | if (error) | |
1327 | putname(s); | |
1328 | else | |
1329 | *name = s; | |
1330 | ||
1331 | return error; | |
1332 | } | |
1333 | ||
1da177e4 LT |
1334 | /* |
1335 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1336 | * minimal. | |
1337 | */ | |
1338 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1339 | { | |
da9592ed DH |
1340 | uid_t fsuid = current_fsuid(); |
1341 | ||
1da177e4 LT |
1342 | if (!(dir->i_mode & S_ISVTX)) |
1343 | return 0; | |
da9592ed | 1344 | if (inode->i_uid == fsuid) |
1da177e4 | 1345 | return 0; |
da9592ed | 1346 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1347 | return 0; |
1348 | return !capable(CAP_FOWNER); | |
1349 | } | |
1350 | ||
1351 | /* | |
1352 | * Check whether we can remove a link victim from directory dir, check | |
1353 | * whether the type of victim is right. | |
1354 | * 1. We can't do it if dir is read-only (done in permission()) | |
1355 | * 2. We should have write and exec permissions on dir | |
1356 | * 3. We can't remove anything from append-only dir | |
1357 | * 4. We can't do anything with immutable dir (done in permission()) | |
1358 | * 5. If the sticky bit on dir is set we should either | |
1359 | * a. be owner of dir, or | |
1360 | * b. be owner of victim, or | |
1361 | * c. have CAP_FOWNER capability | |
1362 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1363 | * links pointing to it. | |
1364 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1365 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1366 | * 9. We can't remove a root or mountpoint. | |
1367 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1368 | * nfs_async_unlink(). | |
1369 | */ | |
858119e1 | 1370 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1371 | { |
1372 | int error; | |
1373 | ||
1374 | if (!victim->d_inode) | |
1375 | return -ENOENT; | |
1376 | ||
1377 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1378 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1379 | |
f419a2e3 | 1380 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1381 | if (error) |
1382 | return error; | |
1383 | if (IS_APPEND(dir)) | |
1384 | return -EPERM; | |
1385 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1386 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1387 | return -EPERM; |
1388 | if (isdir) { | |
1389 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1390 | return -ENOTDIR; | |
1391 | if (IS_ROOT(victim)) | |
1392 | return -EBUSY; | |
1393 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1394 | return -EISDIR; | |
1395 | if (IS_DEADDIR(dir)) | |
1396 | return -ENOENT; | |
1397 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1398 | return -EBUSY; | |
1399 | return 0; | |
1400 | } | |
1401 | ||
1402 | /* Check whether we can create an object with dentry child in directory | |
1403 | * dir. | |
1404 | * 1. We can't do it if child already exists (open has special treatment for | |
1405 | * this case, but since we are inlined it's OK) | |
1406 | * 2. We can't do it if dir is read-only (done in permission()) | |
1407 | * 3. We should have write and exec permissions on dir | |
1408 | * 4. We can't do it if dir is immutable (done in permission()) | |
1409 | */ | |
a95164d9 | 1410 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1411 | { |
1412 | if (child->d_inode) | |
1413 | return -EEXIST; | |
1414 | if (IS_DEADDIR(dir)) | |
1415 | return -ENOENT; | |
f419a2e3 | 1416 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1417 | } |
1418 | ||
1419 | /* | |
1da177e4 LT |
1420 | * O_DIRECTORY translates into forcing a directory lookup. |
1421 | */ | |
1422 | static inline int lookup_flags(unsigned int f) | |
1423 | { | |
1424 | unsigned long retval = LOOKUP_FOLLOW; | |
1425 | ||
1426 | if (f & O_NOFOLLOW) | |
1427 | retval &= ~LOOKUP_FOLLOW; | |
1428 | ||
1da177e4 LT |
1429 | if (f & O_DIRECTORY) |
1430 | retval |= LOOKUP_DIRECTORY; | |
1431 | ||
1432 | return retval; | |
1433 | } | |
1434 | ||
1435 | /* | |
1436 | * p1 and p2 should be directories on the same fs. | |
1437 | */ | |
1438 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1439 | { | |
1440 | struct dentry *p; | |
1441 | ||
1442 | if (p1 == p2) { | |
f2eace23 | 1443 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1444 | return NULL; |
1445 | } | |
1446 | ||
a11f3a05 | 1447 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1448 | |
e2761a11 OH |
1449 | p = d_ancestor(p2, p1); |
1450 | if (p) { | |
1451 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1452 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1453 | return p; | |
1da177e4 LT |
1454 | } |
1455 | ||
e2761a11 OH |
1456 | p = d_ancestor(p1, p2); |
1457 | if (p) { | |
1458 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1459 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1460 | return p; | |
1da177e4 LT |
1461 | } |
1462 | ||
f2eace23 IM |
1463 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1464 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1465 | return NULL; |
1466 | } | |
1467 | ||
1468 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1469 | { | |
1b1dcc1b | 1470 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1471 | if (p1 != p2) { |
1b1dcc1b | 1472 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1473 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1474 | } |
1475 | } | |
1476 | ||
1477 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1478 | struct nameidata *nd) | |
1479 | { | |
a95164d9 | 1480 | int error = may_create(dir, dentry); |
1da177e4 LT |
1481 | |
1482 | if (error) | |
1483 | return error; | |
1484 | ||
1485 | if (!dir->i_op || !dir->i_op->create) | |
1486 | return -EACCES; /* shouldn't it be ENOSYS? */ | |
1487 | mode &= S_IALLUGO; | |
1488 | mode |= S_IFREG; | |
1489 | error = security_inode_create(dir, dentry, mode); | |
1490 | if (error) | |
1491 | return error; | |
1492 | DQUOT_INIT(dir); | |
1493 | error = dir->i_op->create(dir, dentry, mode, nd); | |
a74574aa | 1494 | if (!error) |
f38aa942 | 1495 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1496 | return error; |
1497 | } | |
1498 | ||
1499 | int may_open(struct nameidata *nd, int acc_mode, int flag) | |
1500 | { | |
4ac91378 | 1501 | struct dentry *dentry = nd->path.dentry; |
1da177e4 LT |
1502 | struct inode *inode = dentry->d_inode; |
1503 | int error; | |
1504 | ||
1505 | if (!inode) | |
1506 | return -ENOENT; | |
1507 | ||
1508 | if (S_ISLNK(inode->i_mode)) | |
1509 | return -ELOOP; | |
1510 | ||
974a9f0b | 1511 | if (S_ISDIR(inode->i_mode) && (acc_mode & MAY_WRITE)) |
1da177e4 LT |
1512 | return -EISDIR; |
1513 | ||
1da177e4 LT |
1514 | /* |
1515 | * FIFO's, sockets and device files are special: they don't | |
1516 | * actually live on the filesystem itself, and as such you | |
1517 | * can write to them even if the filesystem is read-only. | |
1518 | */ | |
1519 | if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { | |
1520 | flag &= ~O_TRUNC; | |
1521 | } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) { | |
4ac91378 | 1522 | if (nd->path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1523 | return -EACCES; |
1524 | ||
1525 | flag &= ~O_TRUNC; | |
4a3fd211 | 1526 | } |
b41572e9 DH |
1527 | |
1528 | error = vfs_permission(nd, acc_mode); | |
1529 | if (error) | |
1530 | return error; | |
6146f0d5 MZ |
1531 | |
1532 | error = ima_path_check(&nd->path, | |
1533 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC)); | |
1534 | if (error) | |
1535 | return error; | |
1da177e4 LT |
1536 | /* |
1537 | * An append-only file must be opened in append mode for writing. | |
1538 | */ | |
1539 | if (IS_APPEND(inode)) { | |
1540 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
1541 | return -EPERM; | |
1542 | if (flag & O_TRUNC) | |
1543 | return -EPERM; | |
1544 | } | |
1545 | ||
1546 | /* O_NOATIME can only be set by the owner or superuser */ | |
1547 | if (flag & O_NOATIME) | |
3bd858ab | 1548 | if (!is_owner_or_cap(inode)) |
1da177e4 LT |
1549 | return -EPERM; |
1550 | ||
1551 | /* | |
1552 | * Ensure there are no outstanding leases on the file. | |
1553 | */ | |
1554 | error = break_lease(inode, flag); | |
1555 | if (error) | |
1556 | return error; | |
1557 | ||
1558 | if (flag & O_TRUNC) { | |
1559 | error = get_write_access(inode); | |
1560 | if (error) | |
1561 | return error; | |
1562 | ||
1563 | /* | |
1564 | * Refuse to truncate files with mandatory locks held on them. | |
1565 | */ | |
1566 | error = locks_verify_locked(inode); | |
1567 | if (!error) { | |
1568 | DQUOT_INIT(inode); | |
d139d7ff MS |
1569 | |
1570 | error = do_truncate(dentry, 0, | |
1571 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1572 | NULL); | |
1da177e4 LT |
1573 | } |
1574 | put_write_access(inode); | |
1575 | if (error) | |
1576 | return error; | |
1577 | } else | |
1578 | if (flag & FMODE_WRITE) | |
1579 | DQUOT_INIT(inode); | |
1580 | ||
1581 | return 0; | |
1582 | } | |
1583 | ||
d57999e1 DH |
1584 | /* |
1585 | * Be careful about ever adding any more callers of this | |
1586 | * function. Its flags must be in the namei format, not | |
1587 | * what get passed to sys_open(). | |
1588 | */ | |
1589 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1590 | int flag, int mode) |
1591 | { | |
1592 | int error; | |
4ac91378 | 1593 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1594 | |
1595 | if (!IS_POSIXACL(dir->d_inode)) | |
1596 | mode &= ~current->fs->umask; | |
1597 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); | |
1598 | mutex_unlock(&dir->d_inode->i_mutex); | |
4ac91378 JB |
1599 | dput(nd->path.dentry); |
1600 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1601 | if (error) |
1602 | return error; | |
1603 | /* Don't check for write permission, don't truncate */ | |
1604 | return may_open(nd, 0, flag & ~O_TRUNC); | |
1605 | } | |
1606 | ||
d57999e1 DH |
1607 | /* |
1608 | * Note that while the flag value (low two bits) for sys_open means: | |
1609 | * 00 - read-only | |
1610 | * 01 - write-only | |
1611 | * 10 - read-write | |
1612 | * 11 - special | |
1613 | * it is changed into | |
1614 | * 00 - no permissions needed | |
1615 | * 01 - read-permission | |
1616 | * 10 - write-permission | |
1617 | * 11 - read-write | |
1618 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1619 | * This is more logical, and also allows the 00 "no perm needed" | |
1620 | * to be used for symlinks (where the permissions are checked | |
1621 | * later). | |
1622 | * | |
1623 | */ | |
1624 | static inline int open_to_namei_flags(int flag) | |
1625 | { | |
1626 | if ((flag+1) & O_ACCMODE) | |
1627 | flag++; | |
1628 | return flag; | |
1629 | } | |
1630 | ||
4a3fd211 DH |
1631 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1632 | { | |
1633 | /* | |
1634 | * We'll never write to the fs underlying | |
1635 | * a device file. | |
1636 | */ | |
1637 | if (special_file(inode->i_mode)) | |
1638 | return 0; | |
1639 | return (flag & O_TRUNC); | |
1640 | } | |
1641 | ||
1da177e4 | 1642 | /* |
4a3fd211 DH |
1643 | * Note that the low bits of the passed in "open_flag" |
1644 | * are not the same as in the local variable "flag". See | |
1645 | * open_to_namei_flags() for more details. | |
1da177e4 | 1646 | */ |
a70e65df CH |
1647 | struct file *do_filp_open(int dfd, const char *pathname, |
1648 | int open_flag, int mode) | |
1da177e4 | 1649 | { |
4a3fd211 | 1650 | struct file *filp; |
a70e65df | 1651 | struct nameidata nd; |
834f2a4a | 1652 | int acc_mode, error; |
4e7506e4 | 1653 | struct path path; |
1da177e4 LT |
1654 | struct dentry *dir; |
1655 | int count = 0; | |
4a3fd211 | 1656 | int will_write; |
d57999e1 | 1657 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1658 | |
b77b0646 | 1659 | acc_mode = MAY_OPEN | ACC_MODE(flag); |
1da177e4 | 1660 | |
834f2a4a TM |
1661 | /* O_TRUNC implies we need access checks for write permissions */ |
1662 | if (flag & O_TRUNC) | |
1663 | acc_mode |= MAY_WRITE; | |
1664 | ||
1da177e4 LT |
1665 | /* Allow the LSM permission hook to distinguish append |
1666 | access from general write access. */ | |
1667 | if (flag & O_APPEND) | |
1668 | acc_mode |= MAY_APPEND; | |
1669 | ||
1da177e4 LT |
1670 | /* |
1671 | * The simplest case - just a plain lookup. | |
1672 | */ | |
1673 | if (!(flag & O_CREAT)) { | |
5590ff0d | 1674 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
a70e65df | 1675 | &nd, flag); |
1da177e4 | 1676 | if (error) |
a70e65df | 1677 | return ERR_PTR(error); |
1da177e4 LT |
1678 | goto ok; |
1679 | } | |
1680 | ||
1681 | /* | |
1682 | * Create - we need to know the parent. | |
1683 | */ | |
8737f3a1 | 1684 | error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1685 | if (error) |
a70e65df | 1686 | return ERR_PTR(error); |
1da177e4 LT |
1687 | |
1688 | /* | |
1689 | * We have the parent and last component. First of all, check | |
1690 | * that we are not asked to creat(2) an obvious directory - that | |
1691 | * will not do. | |
1692 | */ | |
1693 | error = -EISDIR; | |
a70e65df | 1694 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1695 | goto exit_parent; |
1da177e4 | 1696 | |
8737f3a1 AV |
1697 | error = -ENFILE; |
1698 | filp = get_empty_filp(); | |
1699 | if (filp == NULL) | |
1700 | goto exit_parent; | |
1701 | nd.intent.open.file = filp; | |
1702 | nd.intent.open.flags = flag; | |
1703 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1704 | dir = nd.path.dentry; |
1705 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1706 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1707 | if (flag & O_EXCL) |
1708 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1709 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1710 | path.dentry = lookup_hash(&nd); |
1711 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1712 | |
1713 | do_last: | |
4e7506e4 AV |
1714 | error = PTR_ERR(path.dentry); |
1715 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1716 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1717 | goto exit; |
1718 | } | |
1719 | ||
a70e65df | 1720 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1721 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1722 | goto exit_mutex_unlock; |
4af4c52f OD |
1723 | } |
1724 | ||
1da177e4 | 1725 | /* Negative dentry, just create the file */ |
4e7506e4 | 1726 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1727 | /* |
1728 | * This write is needed to ensure that a | |
1729 | * ro->rw transition does not occur between | |
1730 | * the time when the file is created and when | |
1731 | * a permanent write count is taken through | |
1732 | * the 'struct file' in nameidata_to_filp(). | |
1733 | */ | |
1734 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1735 | if (error) |
4a3fd211 DH |
1736 | goto exit_mutex_unlock; |
1737 | error = __open_namei_create(&nd, &path, flag, mode); | |
1738 | if (error) { | |
1739 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1740 | goto exit; |
4a3fd211 DH |
1741 | } |
1742 | filp = nameidata_to_filp(&nd, open_flag); | |
1743 | mnt_drop_write(nd.path.mnt); | |
1744 | return filp; | |
1da177e4 LT |
1745 | } |
1746 | ||
1747 | /* | |
1748 | * It already exists. | |
1749 | */ | |
1b1dcc1b | 1750 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1751 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1752 | |
1753 | error = -EEXIST; | |
1754 | if (flag & O_EXCL) | |
1755 | goto exit_dput; | |
1756 | ||
e13b210f | 1757 | if (__follow_mount(&path)) { |
1da177e4 | 1758 | error = -ELOOP; |
ba7a4c1a AV |
1759 | if (flag & O_NOFOLLOW) |
1760 | goto exit_dput; | |
1da177e4 | 1761 | } |
3e2efce0 | 1762 | |
1da177e4 | 1763 | error = -ENOENT; |
4e7506e4 | 1764 | if (!path.dentry->d_inode) |
1da177e4 | 1765 | goto exit_dput; |
4e7506e4 | 1766 | if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1767 | goto do_link; |
1768 | ||
a70e65df | 1769 | path_to_nameidata(&path, &nd); |
1da177e4 | 1770 | error = -EISDIR; |
4e7506e4 | 1771 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1772 | goto exit; |
1773 | ok: | |
4a3fd211 DH |
1774 | /* |
1775 | * Consider: | |
1776 | * 1. may_open() truncates a file | |
1777 | * 2. a rw->ro mount transition occurs | |
1778 | * 3. nameidata_to_filp() fails due to | |
1779 | * the ro mount. | |
1780 | * That would be inconsistent, and should | |
1781 | * be avoided. Taking this mnt write here | |
1782 | * ensures that (2) can not occur. | |
1783 | */ | |
1784 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1785 | if (will_write) { | |
1786 | error = mnt_want_write(nd.path.mnt); | |
1787 | if (error) | |
1788 | goto exit; | |
1789 | } | |
a70e65df | 1790 | error = may_open(&nd, acc_mode, flag); |
4a3fd211 DH |
1791 | if (error) { |
1792 | if (will_write) | |
1793 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1794 | goto exit; |
4a3fd211 DH |
1795 | } |
1796 | filp = nameidata_to_filp(&nd, open_flag); | |
1797 | /* | |
1798 | * It is now safe to drop the mnt write | |
1799 | * because the filp has had a write taken | |
1800 | * on its behalf. | |
1801 | */ | |
1802 | if (will_write) | |
1803 | mnt_drop_write(nd.path.mnt); | |
1804 | return filp; | |
1da177e4 | 1805 | |
4a3fd211 DH |
1806 | exit_mutex_unlock: |
1807 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1808 | exit_dput: |
a70e65df | 1809 | path_put_conditional(&path, &nd); |
1da177e4 | 1810 | exit: |
a70e65df CH |
1811 | if (!IS_ERR(nd.intent.open.file)) |
1812 | release_open_intent(&nd); | |
8737f3a1 | 1813 | exit_parent: |
a70e65df CH |
1814 | path_put(&nd.path); |
1815 | return ERR_PTR(error); | |
1da177e4 LT |
1816 | |
1817 | do_link: | |
1818 | error = -ELOOP; | |
1819 | if (flag & O_NOFOLLOW) | |
1820 | goto exit_dput; | |
1821 | /* | |
1822 | * This is subtle. Instead of calling do_follow_link() we do the | |
1823 | * thing by hands. The reason is that this way we have zero link_count | |
1824 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1825 | * After that we have the parent and last component, i.e. | |
1826 | * we are in the same situation as after the first path_walk(). | |
1827 | * Well, almost - if the last component is normal we get its copy | |
1828 | * stored in nd->last.name and we will have to putname() it when we | |
1829 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1830 | */ | |
a70e65df CH |
1831 | nd.flags |= LOOKUP_PARENT; |
1832 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1833 | if (error) |
1834 | goto exit_dput; | |
a70e65df | 1835 | error = __do_follow_link(&path, &nd); |
de459215 KK |
1836 | if (error) { |
1837 | /* Does someone understand code flow here? Or it is only | |
1838 | * me so stupid? Anathema to whoever designed this non-sense | |
1839 | * with "intent.open". | |
1840 | */ | |
a70e65df CH |
1841 | release_open_intent(&nd); |
1842 | return ERR_PTR(error); | |
de459215 | 1843 | } |
a70e65df CH |
1844 | nd.flags &= ~LOOKUP_PARENT; |
1845 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1846 | goto ok; |
1da177e4 | 1847 | error = -EISDIR; |
a70e65df | 1848 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1849 | goto exit; |
a70e65df CH |
1850 | if (nd.last.name[nd.last.len]) { |
1851 | __putname(nd.last.name); | |
1da177e4 LT |
1852 | goto exit; |
1853 | } | |
1854 | error = -ELOOP; | |
1855 | if (count++==32) { | |
a70e65df | 1856 | __putname(nd.last.name); |
1da177e4 LT |
1857 | goto exit; |
1858 | } | |
a70e65df | 1859 | dir = nd.path.dentry; |
1b1dcc1b | 1860 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1861 | path.dentry = lookup_hash(&nd); |
1862 | path.mnt = nd.path.mnt; | |
1863 | __putname(nd.last.name); | |
1da177e4 LT |
1864 | goto do_last; |
1865 | } | |
1866 | ||
a70e65df CH |
1867 | /** |
1868 | * filp_open - open file and return file pointer | |
1869 | * | |
1870 | * @filename: path to open | |
1871 | * @flags: open flags as per the open(2) second argument | |
1872 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1873 | * | |
1874 | * This is the helper to open a file from kernelspace if you really | |
1875 | * have to. But in generally you should not do this, so please move | |
1876 | * along, nothing to see here.. | |
1877 | */ | |
1878 | struct file *filp_open(const char *filename, int flags, int mode) | |
1879 | { | |
1880 | return do_filp_open(AT_FDCWD, filename, flags, mode); | |
1881 | } | |
1882 | EXPORT_SYMBOL(filp_open); | |
1883 | ||
1da177e4 LT |
1884 | /** |
1885 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1886 | * @nd: nameidata info | |
1887 | * @is_dir: directory flag | |
1888 | * | |
1889 | * Simple function to lookup and return a dentry and create it | |
1890 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1891 | * |
4ac91378 | 1892 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1893 | */ |
1894 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1895 | { | |
c663e5d8 | 1896 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1897 | |
4ac91378 | 1898 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1899 | /* |
1900 | * Yucky last component or no last component at all? | |
1901 | * (foo/., foo/.., /////) | |
1902 | */ | |
1da177e4 LT |
1903 | if (nd->last_type != LAST_NORM) |
1904 | goto fail; | |
1905 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1906 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1907 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1908 | |
1909 | /* | |
1910 | * Do the final lookup. | |
1911 | */ | |
49705b77 | 1912 | dentry = lookup_hash(nd); |
1da177e4 LT |
1913 | if (IS_ERR(dentry)) |
1914 | goto fail; | |
c663e5d8 | 1915 | |
e9baf6e5 AV |
1916 | if (dentry->d_inode) |
1917 | goto eexist; | |
c663e5d8 CH |
1918 | /* |
1919 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1920 | * From the vfs_mknod() POV we just have a negative dentry - | |
1921 | * all is fine. Let's be bastards - you had / on the end, you've | |
1922 | * been asking for (non-existent) directory. -ENOENT for you. | |
1923 | */ | |
e9baf6e5 AV |
1924 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1925 | dput(dentry); | |
1926 | dentry = ERR_PTR(-ENOENT); | |
1927 | } | |
1da177e4 | 1928 | return dentry; |
e9baf6e5 | 1929 | eexist: |
1da177e4 | 1930 | dput(dentry); |
e9baf6e5 | 1931 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1932 | fail: |
1933 | return dentry; | |
1934 | } | |
f81a0bff | 1935 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1936 | |
1937 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1938 | { | |
a95164d9 | 1939 | int error = may_create(dir, dentry); |
1da177e4 LT |
1940 | |
1941 | if (error) | |
1942 | return error; | |
1943 | ||
1944 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1945 | return -EPERM; | |
1946 | ||
1947 | if (!dir->i_op || !dir->i_op->mknod) | |
1948 | return -EPERM; | |
1949 | ||
08ce5f16 SH |
1950 | error = devcgroup_inode_mknod(mode, dev); |
1951 | if (error) | |
1952 | return error; | |
1953 | ||
1da177e4 LT |
1954 | error = security_inode_mknod(dir, dentry, mode, dev); |
1955 | if (error) | |
1956 | return error; | |
1957 | ||
1958 | DQUOT_INIT(dir); | |
1959 | error = dir->i_op->mknod(dir, dentry, mode, dev); | |
a74574aa | 1960 | if (!error) |
f38aa942 | 1961 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1962 | return error; |
1963 | } | |
1964 | ||
463c3197 DH |
1965 | static int may_mknod(mode_t mode) |
1966 | { | |
1967 | switch (mode & S_IFMT) { | |
1968 | case S_IFREG: | |
1969 | case S_IFCHR: | |
1970 | case S_IFBLK: | |
1971 | case S_IFIFO: | |
1972 | case S_IFSOCK: | |
1973 | case 0: /* zero mode translates to S_IFREG */ | |
1974 | return 0; | |
1975 | case S_IFDIR: | |
1976 | return -EPERM; | |
1977 | default: | |
1978 | return -EINVAL; | |
1979 | } | |
1980 | } | |
1981 | ||
5590ff0d UD |
1982 | asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode, |
1983 | unsigned dev) | |
1da177e4 | 1984 | { |
2ad94ae6 AV |
1985 | int error; |
1986 | char *tmp; | |
1987 | struct dentry *dentry; | |
1da177e4 LT |
1988 | struct nameidata nd; |
1989 | ||
1990 | if (S_ISDIR(mode)) | |
1991 | return -EPERM; | |
1da177e4 | 1992 | |
2ad94ae6 | 1993 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 1994 | if (error) |
2ad94ae6 AV |
1995 | return error; |
1996 | ||
1da177e4 | 1997 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
1998 | if (IS_ERR(dentry)) { |
1999 | error = PTR_ERR(dentry); | |
2000 | goto out_unlock; | |
2001 | } | |
4ac91378 | 2002 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
1da177e4 | 2003 | mode &= ~current->fs->umask; |
463c3197 DH |
2004 | error = may_mknod(mode); |
2005 | if (error) | |
2006 | goto out_dput; | |
2007 | error = mnt_want_write(nd.path.mnt); | |
2008 | if (error) | |
2009 | goto out_dput; | |
2010 | switch (mode & S_IFMT) { | |
1da177e4 | 2011 | case 0: case S_IFREG: |
4ac91378 | 2012 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2013 | break; |
2014 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2015 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2016 | new_decode_dev(dev)); |
2017 | break; | |
2018 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2019 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2020 | break; |
1da177e4 | 2021 | } |
463c3197 DH |
2022 | mnt_drop_write(nd.path.mnt); |
2023 | out_dput: | |
2024 | dput(dentry); | |
2025 | out_unlock: | |
4ac91378 | 2026 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2027 | path_put(&nd.path); |
1da177e4 LT |
2028 | putname(tmp); |
2029 | ||
2030 | return error; | |
2031 | } | |
2032 | ||
5590ff0d UD |
2033 | asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev) |
2034 | { | |
2035 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2036 | } | |
2037 | ||
1da177e4 LT |
2038 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2039 | { | |
a95164d9 | 2040 | int error = may_create(dir, dentry); |
1da177e4 LT |
2041 | |
2042 | if (error) | |
2043 | return error; | |
2044 | ||
2045 | if (!dir->i_op || !dir->i_op->mkdir) | |
2046 | return -EPERM; | |
2047 | ||
2048 | mode &= (S_IRWXUGO|S_ISVTX); | |
2049 | error = security_inode_mkdir(dir, dentry, mode); | |
2050 | if (error) | |
2051 | return error; | |
2052 | ||
2053 | DQUOT_INIT(dir); | |
2054 | error = dir->i_op->mkdir(dir, dentry, mode); | |
a74574aa | 2055 | if (!error) |
f38aa942 | 2056 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2057 | return error; |
2058 | } | |
2059 | ||
5590ff0d | 2060 | asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode) |
1da177e4 LT |
2061 | { |
2062 | int error = 0; | |
2063 | char * tmp; | |
6902d925 DH |
2064 | struct dentry *dentry; |
2065 | struct nameidata nd; | |
1da177e4 | 2066 | |
2ad94ae6 AV |
2067 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2068 | if (error) | |
6902d925 | 2069 | goto out_err; |
1da177e4 | 2070 | |
6902d925 DH |
2071 | dentry = lookup_create(&nd, 1); |
2072 | error = PTR_ERR(dentry); | |
2073 | if (IS_ERR(dentry)) | |
2074 | goto out_unlock; | |
1da177e4 | 2075 | |
4ac91378 | 2076 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
6902d925 | 2077 | mode &= ~current->fs->umask; |
463c3197 DH |
2078 | error = mnt_want_write(nd.path.mnt); |
2079 | if (error) | |
2080 | goto out_dput; | |
4ac91378 | 2081 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
463c3197 DH |
2082 | mnt_drop_write(nd.path.mnt); |
2083 | out_dput: | |
6902d925 DH |
2084 | dput(dentry); |
2085 | out_unlock: | |
4ac91378 | 2086 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2087 | path_put(&nd.path); |
6902d925 DH |
2088 | putname(tmp); |
2089 | out_err: | |
1da177e4 LT |
2090 | return error; |
2091 | } | |
2092 | ||
5590ff0d UD |
2093 | asmlinkage long sys_mkdir(const char __user *pathname, int mode) |
2094 | { | |
2095 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2096 | } | |
2097 | ||
1da177e4 LT |
2098 | /* |
2099 | * We try to drop the dentry early: we should have | |
2100 | * a usage count of 2 if we're the only user of this | |
2101 | * dentry, and if that is true (possibly after pruning | |
2102 | * the dcache), then we drop the dentry now. | |
2103 | * | |
2104 | * A low-level filesystem can, if it choses, legally | |
2105 | * do a | |
2106 | * | |
2107 | * if (!d_unhashed(dentry)) | |
2108 | * return -EBUSY; | |
2109 | * | |
2110 | * if it cannot handle the case of removing a directory | |
2111 | * that is still in use by something else.. | |
2112 | */ | |
2113 | void dentry_unhash(struct dentry *dentry) | |
2114 | { | |
2115 | dget(dentry); | |
dc168427 | 2116 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2117 | spin_lock(&dcache_lock); |
2118 | spin_lock(&dentry->d_lock); | |
2119 | if (atomic_read(&dentry->d_count) == 2) | |
2120 | __d_drop(dentry); | |
2121 | spin_unlock(&dentry->d_lock); | |
2122 | spin_unlock(&dcache_lock); | |
2123 | } | |
2124 | ||
2125 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2126 | { | |
2127 | int error = may_delete(dir, dentry, 1); | |
2128 | ||
2129 | if (error) | |
2130 | return error; | |
2131 | ||
2132 | if (!dir->i_op || !dir->i_op->rmdir) | |
2133 | return -EPERM; | |
2134 | ||
2135 | DQUOT_INIT(dir); | |
2136 | ||
1b1dcc1b | 2137 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2138 | dentry_unhash(dentry); |
2139 | if (d_mountpoint(dentry)) | |
2140 | error = -EBUSY; | |
2141 | else { | |
2142 | error = security_inode_rmdir(dir, dentry); | |
2143 | if (!error) { | |
2144 | error = dir->i_op->rmdir(dir, dentry); | |
2145 | if (!error) | |
2146 | dentry->d_inode->i_flags |= S_DEAD; | |
2147 | } | |
2148 | } | |
1b1dcc1b | 2149 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2150 | if (!error) { |
1da177e4 LT |
2151 | d_delete(dentry); |
2152 | } | |
2153 | dput(dentry); | |
2154 | ||
2155 | return error; | |
2156 | } | |
2157 | ||
5590ff0d | 2158 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2159 | { |
2160 | int error = 0; | |
2161 | char * name; | |
2162 | struct dentry *dentry; | |
2163 | struct nameidata nd; | |
2164 | ||
2ad94ae6 | 2165 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2166 | if (error) |
2ad94ae6 | 2167 | return error; |
1da177e4 LT |
2168 | |
2169 | switch(nd.last_type) { | |
0612d9fb OH |
2170 | case LAST_DOTDOT: |
2171 | error = -ENOTEMPTY; | |
2172 | goto exit1; | |
2173 | case LAST_DOT: | |
2174 | error = -EINVAL; | |
2175 | goto exit1; | |
2176 | case LAST_ROOT: | |
2177 | error = -EBUSY; | |
2178 | goto exit1; | |
1da177e4 | 2179 | } |
0612d9fb OH |
2180 | |
2181 | nd.flags &= ~LOOKUP_PARENT; | |
2182 | ||
4ac91378 | 2183 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2184 | dentry = lookup_hash(&nd); |
1da177e4 | 2185 | error = PTR_ERR(dentry); |
6902d925 DH |
2186 | if (IS_ERR(dentry)) |
2187 | goto exit2; | |
0622753b DH |
2188 | error = mnt_want_write(nd.path.mnt); |
2189 | if (error) | |
2190 | goto exit3; | |
4ac91378 | 2191 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
0622753b DH |
2192 | mnt_drop_write(nd.path.mnt); |
2193 | exit3: | |
6902d925 DH |
2194 | dput(dentry); |
2195 | exit2: | |
4ac91378 | 2196 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2197 | exit1: |
1d957f9b | 2198 | path_put(&nd.path); |
1da177e4 LT |
2199 | putname(name); |
2200 | return error; | |
2201 | } | |
2202 | ||
5590ff0d UD |
2203 | asmlinkage long sys_rmdir(const char __user *pathname) |
2204 | { | |
2205 | return do_rmdir(AT_FDCWD, pathname); | |
2206 | } | |
2207 | ||
1da177e4 LT |
2208 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2209 | { | |
2210 | int error = may_delete(dir, dentry, 0); | |
2211 | ||
2212 | if (error) | |
2213 | return error; | |
2214 | ||
2215 | if (!dir->i_op || !dir->i_op->unlink) | |
2216 | return -EPERM; | |
2217 | ||
2218 | DQUOT_INIT(dir); | |
2219 | ||
1b1dcc1b | 2220 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2221 | if (d_mountpoint(dentry)) |
2222 | error = -EBUSY; | |
2223 | else { | |
2224 | error = security_inode_unlink(dir, dentry); | |
2225 | if (!error) | |
2226 | error = dir->i_op->unlink(dir, dentry); | |
2227 | } | |
1b1dcc1b | 2228 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2229 | |
2230 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2231 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2232 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2233 | d_delete(dentry); |
1da177e4 | 2234 | } |
0eeca283 | 2235 | |
1da177e4 LT |
2236 | return error; |
2237 | } | |
2238 | ||
2239 | /* | |
2240 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2241 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2242 | * writeout happening, and we don't want to prevent access to the directory |
2243 | * while waiting on the I/O. | |
2244 | */ | |
5590ff0d | 2245 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2246 | { |
2ad94ae6 AV |
2247 | int error; |
2248 | char *name; | |
1da177e4 LT |
2249 | struct dentry *dentry; |
2250 | struct nameidata nd; | |
2251 | struct inode *inode = NULL; | |
2252 | ||
2ad94ae6 | 2253 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2254 | if (error) |
2ad94ae6 AV |
2255 | return error; |
2256 | ||
1da177e4 LT |
2257 | error = -EISDIR; |
2258 | if (nd.last_type != LAST_NORM) | |
2259 | goto exit1; | |
0612d9fb OH |
2260 | |
2261 | nd.flags &= ~LOOKUP_PARENT; | |
2262 | ||
4ac91378 | 2263 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2264 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2265 | error = PTR_ERR(dentry); |
2266 | if (!IS_ERR(dentry)) { | |
2267 | /* Why not before? Because we want correct error value */ | |
2268 | if (nd.last.name[nd.last.len]) | |
2269 | goto slashes; | |
2270 | inode = dentry->d_inode; | |
2271 | if (inode) | |
2272 | atomic_inc(&inode->i_count); | |
0622753b DH |
2273 | error = mnt_want_write(nd.path.mnt); |
2274 | if (error) | |
2275 | goto exit2; | |
4ac91378 | 2276 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
0622753b | 2277 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2278 | exit2: |
2279 | dput(dentry); | |
2280 | } | |
4ac91378 | 2281 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2282 | if (inode) |
2283 | iput(inode); /* truncate the inode here */ | |
2284 | exit1: | |
1d957f9b | 2285 | path_put(&nd.path); |
1da177e4 LT |
2286 | putname(name); |
2287 | return error; | |
2288 | ||
2289 | slashes: | |
2290 | error = !dentry->d_inode ? -ENOENT : | |
2291 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2292 | goto exit2; | |
2293 | } | |
2294 | ||
5590ff0d UD |
2295 | asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag) |
2296 | { | |
2297 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2298 | return -EINVAL; | |
2299 | ||
2300 | if (flag & AT_REMOVEDIR) | |
2301 | return do_rmdir(dfd, pathname); | |
2302 | ||
2303 | return do_unlinkat(dfd, pathname); | |
2304 | } | |
2305 | ||
2306 | asmlinkage long sys_unlink(const char __user *pathname) | |
2307 | { | |
2308 | return do_unlinkat(AT_FDCWD, pathname); | |
2309 | } | |
2310 | ||
db2e747b | 2311 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2312 | { |
a95164d9 | 2313 | int error = may_create(dir, dentry); |
1da177e4 LT |
2314 | |
2315 | if (error) | |
2316 | return error; | |
2317 | ||
2318 | if (!dir->i_op || !dir->i_op->symlink) | |
2319 | return -EPERM; | |
2320 | ||
2321 | error = security_inode_symlink(dir, dentry, oldname); | |
2322 | if (error) | |
2323 | return error; | |
2324 | ||
2325 | DQUOT_INIT(dir); | |
2326 | error = dir->i_op->symlink(dir, dentry, oldname); | |
a74574aa | 2327 | if (!error) |
f38aa942 | 2328 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2329 | return error; |
2330 | } | |
2331 | ||
5590ff0d UD |
2332 | asmlinkage long sys_symlinkat(const char __user *oldname, |
2333 | int newdfd, const char __user *newname) | |
1da177e4 | 2334 | { |
2ad94ae6 AV |
2335 | int error; |
2336 | char *from; | |
2337 | char *to; | |
6902d925 DH |
2338 | struct dentry *dentry; |
2339 | struct nameidata nd; | |
1da177e4 LT |
2340 | |
2341 | from = getname(oldname); | |
2ad94ae6 | 2342 | if (IS_ERR(from)) |
1da177e4 | 2343 | return PTR_ERR(from); |
1da177e4 | 2344 | |
2ad94ae6 | 2345 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2346 | if (error) |
2ad94ae6 AV |
2347 | goto out_putname; |
2348 | ||
6902d925 DH |
2349 | dentry = lookup_create(&nd, 0); |
2350 | error = PTR_ERR(dentry); | |
2351 | if (IS_ERR(dentry)) | |
2352 | goto out_unlock; | |
2353 | ||
75c3f29d DH |
2354 | error = mnt_want_write(nd.path.mnt); |
2355 | if (error) | |
2356 | goto out_dput; | |
db2e747b | 2357 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
75c3f29d DH |
2358 | mnt_drop_write(nd.path.mnt); |
2359 | out_dput: | |
6902d925 DH |
2360 | dput(dentry); |
2361 | out_unlock: | |
4ac91378 | 2362 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2363 | path_put(&nd.path); |
6902d925 DH |
2364 | putname(to); |
2365 | out_putname: | |
1da177e4 LT |
2366 | putname(from); |
2367 | return error; | |
2368 | } | |
2369 | ||
5590ff0d UD |
2370 | asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname) |
2371 | { | |
2372 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2373 | } | |
2374 | ||
1da177e4 LT |
2375 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2376 | { | |
2377 | struct inode *inode = old_dentry->d_inode; | |
2378 | int error; | |
2379 | ||
2380 | if (!inode) | |
2381 | return -ENOENT; | |
2382 | ||
a95164d9 | 2383 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2384 | if (error) |
2385 | return error; | |
2386 | ||
2387 | if (dir->i_sb != inode->i_sb) | |
2388 | return -EXDEV; | |
2389 | ||
2390 | /* | |
2391 | * A link to an append-only or immutable file cannot be created. | |
2392 | */ | |
2393 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2394 | return -EPERM; | |
2395 | if (!dir->i_op || !dir->i_op->link) | |
2396 | return -EPERM; | |
7e79eedb | 2397 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2398 | return -EPERM; |
2399 | ||
2400 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2401 | if (error) | |
2402 | return error; | |
2403 | ||
7e79eedb | 2404 | mutex_lock(&inode->i_mutex); |
1da177e4 LT |
2405 | DQUOT_INIT(dir); |
2406 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 2407 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2408 | if (!error) |
7e79eedb | 2409 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2410 | return error; |
2411 | } | |
2412 | ||
2413 | /* | |
2414 | * Hardlinks are often used in delicate situations. We avoid | |
2415 | * security-related surprises by not following symlinks on the | |
2416 | * newname. --KAB | |
2417 | * | |
2418 | * We don't follow them on the oldname either to be compatible | |
2419 | * with linux 2.0, and to avoid hard-linking to directories | |
2420 | * and other special files. --ADM | |
2421 | */ | |
5590ff0d | 2422 | asmlinkage long sys_linkat(int olddfd, const char __user *oldname, |
c04030e1 UD |
2423 | int newdfd, const char __user *newname, |
2424 | int flags) | |
1da177e4 LT |
2425 | { |
2426 | struct dentry *new_dentry; | |
2d8f3038 AV |
2427 | struct nameidata nd; |
2428 | struct path old_path; | |
1da177e4 | 2429 | int error; |
2ad94ae6 | 2430 | char *to; |
1da177e4 | 2431 | |
45c9b11a | 2432 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2433 | return -EINVAL; |
2434 | ||
2d8f3038 AV |
2435 | error = user_path_at(olddfd, oldname, |
2436 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2437 | &old_path); | |
1da177e4 | 2438 | if (error) |
2ad94ae6 AV |
2439 | return error; |
2440 | ||
2441 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2442 | if (error) |
2443 | goto out; | |
2444 | error = -EXDEV; | |
2d8f3038 | 2445 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2446 | goto out_release; |
2447 | new_dentry = lookup_create(&nd, 0); | |
2448 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2449 | if (IS_ERR(new_dentry)) |
2450 | goto out_unlock; | |
75c3f29d DH |
2451 | error = mnt_want_write(nd.path.mnt); |
2452 | if (error) | |
2453 | goto out_dput; | |
2d8f3038 | 2454 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
75c3f29d DH |
2455 | mnt_drop_write(nd.path.mnt); |
2456 | out_dput: | |
6902d925 DH |
2457 | dput(new_dentry); |
2458 | out_unlock: | |
4ac91378 | 2459 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2460 | out_release: |
1d957f9b | 2461 | path_put(&nd.path); |
2ad94ae6 | 2462 | putname(to); |
1da177e4 | 2463 | out: |
2d8f3038 | 2464 | path_put(&old_path); |
1da177e4 LT |
2465 | |
2466 | return error; | |
2467 | } | |
2468 | ||
5590ff0d UD |
2469 | asmlinkage long sys_link(const char __user *oldname, const char __user *newname) |
2470 | { | |
c04030e1 | 2471 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2472 | } |
2473 | ||
1da177e4 LT |
2474 | /* |
2475 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2476 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2477 | * Problems: | |
2478 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2479 | * b) race potential - two innocent renames can create a loop together. | |
2480 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2481 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2482 | * story. |
2483 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2484 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2485 | * whether the target exists). Solution: try to be smart with locking |
2486 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2487 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2488 | * move will be locked. Thus we can rank directories by the tree |
2489 | * (ancestors first) and rank all non-directories after them. | |
2490 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2491 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2492 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2493 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2494 | * we'd better make sure that there's no link(2) for them. | |
2495 | * d) some filesystems don't support opened-but-unlinked directories, | |
2496 | * either because of layout or because they are not ready to deal with | |
2497 | * all cases correctly. The latter will be fixed (taking this sort of | |
2498 | * stuff into VFS), but the former is not going away. Solution: the same | |
2499 | * trick as in rmdir(). | |
2500 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2501 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2502 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2503 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2504 | * locking]. |
2505 | */ | |
75c96f85 AB |
2506 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2507 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2508 | { |
2509 | int error = 0; | |
2510 | struct inode *target; | |
2511 | ||
2512 | /* | |
2513 | * If we are going to change the parent - check write permissions, | |
2514 | * we'll need to flip '..'. | |
2515 | */ | |
2516 | if (new_dir != old_dir) { | |
f419a2e3 | 2517 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2518 | if (error) |
2519 | return error; | |
2520 | } | |
2521 | ||
2522 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2523 | if (error) | |
2524 | return error; | |
2525 | ||
2526 | target = new_dentry->d_inode; | |
2527 | if (target) { | |
1b1dcc1b | 2528 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2529 | dentry_unhash(new_dentry); |
2530 | } | |
2531 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2532 | error = -EBUSY; | |
2533 | else | |
2534 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2535 | if (target) { | |
2536 | if (!error) | |
2537 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2538 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2539 | if (d_unhashed(new_dentry)) |
2540 | d_rehash(new_dentry); | |
2541 | dput(new_dentry); | |
2542 | } | |
e31e14ec | 2543 | if (!error) |
349457cc MF |
2544 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2545 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2546 | return error; |
2547 | } | |
2548 | ||
75c96f85 AB |
2549 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2550 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2551 | { |
2552 | struct inode *target; | |
2553 | int error; | |
2554 | ||
2555 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2556 | if (error) | |
2557 | return error; | |
2558 | ||
2559 | dget(new_dentry); | |
2560 | target = new_dentry->d_inode; | |
2561 | if (target) | |
1b1dcc1b | 2562 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2563 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2564 | error = -EBUSY; | |
2565 | else | |
2566 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2567 | if (!error) { | |
349457cc | 2568 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2569 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2570 | } |
2571 | if (target) | |
1b1dcc1b | 2572 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2573 | dput(new_dentry); |
2574 | return error; | |
2575 | } | |
2576 | ||
2577 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2578 | struct inode *new_dir, struct dentry *new_dentry) | |
2579 | { | |
2580 | int error; | |
2581 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2582 | const char *old_name; |
1da177e4 LT |
2583 | |
2584 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2585 | return 0; | |
2586 | ||
2587 | error = may_delete(old_dir, old_dentry, is_dir); | |
2588 | if (error) | |
2589 | return error; | |
2590 | ||
2591 | if (!new_dentry->d_inode) | |
a95164d9 | 2592 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2593 | else |
2594 | error = may_delete(new_dir, new_dentry, is_dir); | |
2595 | if (error) | |
2596 | return error; | |
2597 | ||
2598 | if (!old_dir->i_op || !old_dir->i_op->rename) | |
2599 | return -EPERM; | |
2600 | ||
2601 | DQUOT_INIT(old_dir); | |
2602 | DQUOT_INIT(new_dir); | |
2603 | ||
0eeca283 RL |
2604 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2605 | ||
1da177e4 LT |
2606 | if (is_dir) |
2607 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2608 | else | |
2609 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2610 | if (!error) { | |
0eeca283 | 2611 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2612 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2613 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2614 | } |
0eeca283 RL |
2615 | fsnotify_oldname_free(old_name); |
2616 | ||
1da177e4 LT |
2617 | return error; |
2618 | } | |
2619 | ||
2ad94ae6 AV |
2620 | asmlinkage long sys_renameat(int olddfd, const char __user *oldname, |
2621 | int newdfd, const char __user *newname) | |
1da177e4 | 2622 | { |
2ad94ae6 AV |
2623 | struct dentry *old_dir, *new_dir; |
2624 | struct dentry *old_dentry, *new_dentry; | |
2625 | struct dentry *trap; | |
1da177e4 | 2626 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2627 | char *from; |
2628 | char *to; | |
2629 | int error; | |
1da177e4 | 2630 | |
2ad94ae6 | 2631 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2632 | if (error) |
2633 | goto exit; | |
2634 | ||
2ad94ae6 | 2635 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2636 | if (error) |
2637 | goto exit1; | |
2638 | ||
2639 | error = -EXDEV; | |
4ac91378 | 2640 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2641 | goto exit2; |
2642 | ||
4ac91378 | 2643 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2644 | error = -EBUSY; |
2645 | if (oldnd.last_type != LAST_NORM) | |
2646 | goto exit2; | |
2647 | ||
4ac91378 | 2648 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2649 | if (newnd.last_type != LAST_NORM) |
2650 | goto exit2; | |
2651 | ||
0612d9fb OH |
2652 | oldnd.flags &= ~LOOKUP_PARENT; |
2653 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2654 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2655 | |
1da177e4 LT |
2656 | trap = lock_rename(new_dir, old_dir); |
2657 | ||
49705b77 | 2658 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2659 | error = PTR_ERR(old_dentry); |
2660 | if (IS_ERR(old_dentry)) | |
2661 | goto exit3; | |
2662 | /* source must exist */ | |
2663 | error = -ENOENT; | |
2664 | if (!old_dentry->d_inode) | |
2665 | goto exit4; | |
2666 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2667 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2668 | error = -ENOTDIR; | |
2669 | if (oldnd.last.name[oldnd.last.len]) | |
2670 | goto exit4; | |
2671 | if (newnd.last.name[newnd.last.len]) | |
2672 | goto exit4; | |
2673 | } | |
2674 | /* source should not be ancestor of target */ | |
2675 | error = -EINVAL; | |
2676 | if (old_dentry == trap) | |
2677 | goto exit4; | |
49705b77 | 2678 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2679 | error = PTR_ERR(new_dentry); |
2680 | if (IS_ERR(new_dentry)) | |
2681 | goto exit4; | |
2682 | /* target should not be an ancestor of source */ | |
2683 | error = -ENOTEMPTY; | |
2684 | if (new_dentry == trap) | |
2685 | goto exit5; | |
2686 | ||
9079b1eb DH |
2687 | error = mnt_want_write(oldnd.path.mnt); |
2688 | if (error) | |
2689 | goto exit5; | |
1da177e4 LT |
2690 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2691 | new_dir->d_inode, new_dentry); | |
9079b1eb | 2692 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2693 | exit5: |
2694 | dput(new_dentry); | |
2695 | exit4: | |
2696 | dput(old_dentry); | |
2697 | exit3: | |
2698 | unlock_rename(new_dir, old_dir); | |
2699 | exit2: | |
1d957f9b | 2700 | path_put(&newnd.path); |
2ad94ae6 | 2701 | putname(to); |
1da177e4 | 2702 | exit1: |
1d957f9b | 2703 | path_put(&oldnd.path); |
1da177e4 | 2704 | putname(from); |
2ad94ae6 | 2705 | exit: |
1da177e4 LT |
2706 | return error; |
2707 | } | |
2708 | ||
5590ff0d UD |
2709 | asmlinkage long sys_rename(const char __user *oldname, const char __user *newname) |
2710 | { | |
2711 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2712 | } | |
2713 | ||
1da177e4 LT |
2714 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2715 | { | |
2716 | int len; | |
2717 | ||
2718 | len = PTR_ERR(link); | |
2719 | if (IS_ERR(link)) | |
2720 | goto out; | |
2721 | ||
2722 | len = strlen(link); | |
2723 | if (len > (unsigned) buflen) | |
2724 | len = buflen; | |
2725 | if (copy_to_user(buffer, link, len)) | |
2726 | len = -EFAULT; | |
2727 | out: | |
2728 | return len; | |
2729 | } | |
2730 | ||
2731 | /* | |
2732 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2733 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2734 | * using) it for any given inode is up to filesystem. | |
2735 | */ | |
2736 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2737 | { | |
2738 | struct nameidata nd; | |
cc314eef | 2739 | void *cookie; |
694a1764 | 2740 | int res; |
cc314eef | 2741 | |
1da177e4 | 2742 | nd.depth = 0; |
cc314eef | 2743 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2744 | if (IS_ERR(cookie)) |
2745 | return PTR_ERR(cookie); | |
2746 | ||
2747 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2748 | if (dentry->d_inode->i_op->put_link) | |
2749 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2750 | return res; | |
1da177e4 LT |
2751 | } |
2752 | ||
2753 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2754 | { | |
2755 | return __vfs_follow_link(nd, link); | |
2756 | } | |
2757 | ||
2758 | /* get the link contents into pagecache */ | |
2759 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2760 | { | |
2761 | struct page * page; | |
2762 | struct address_space *mapping = dentry->d_inode->i_mapping; | |
090d2b18 | 2763 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2764 | if (IS_ERR(page)) |
6fe6900e | 2765 | return (char*)page; |
1da177e4 LT |
2766 | *ppage = page; |
2767 | return kmap(page); | |
1da177e4 LT |
2768 | } |
2769 | ||
2770 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2771 | { | |
2772 | struct page *page = NULL; | |
2773 | char *s = page_getlink(dentry, &page); | |
2774 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2775 | if (page) { | |
2776 | kunmap(page); | |
2777 | page_cache_release(page); | |
2778 | } | |
2779 | return res; | |
2780 | } | |
2781 | ||
cc314eef | 2782 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2783 | { |
cc314eef | 2784 | struct page *page = NULL; |
1da177e4 | 2785 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2786 | return page; |
1da177e4 LT |
2787 | } |
2788 | ||
cc314eef | 2789 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2790 | { |
cc314eef LT |
2791 | struct page *page = cookie; |
2792 | ||
2793 | if (page) { | |
1da177e4 LT |
2794 | kunmap(page); |
2795 | page_cache_release(page); | |
1da177e4 LT |
2796 | } |
2797 | } | |
2798 | ||
0adb25d2 KK |
2799 | int __page_symlink(struct inode *inode, const char *symname, int len, |
2800 | gfp_t gfp_mask) | |
1da177e4 LT |
2801 | { |
2802 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2803 | struct page *page; |
afddba49 | 2804 | void *fsdata; |
beb497ab | 2805 | int err; |
1da177e4 LT |
2806 | char *kaddr; |
2807 | ||
7e53cac4 | 2808 | retry: |
afddba49 NP |
2809 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
2810 | AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata); | |
1da177e4 | 2811 | if (err) |
afddba49 NP |
2812 | goto fail; |
2813 | ||
1da177e4 LT |
2814 | kaddr = kmap_atomic(page, KM_USER0); |
2815 | memcpy(kaddr, symname, len-1); | |
2816 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2817 | |
2818 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2819 | page, fsdata); | |
1da177e4 LT |
2820 | if (err < 0) |
2821 | goto fail; | |
afddba49 NP |
2822 | if (err < len-1) |
2823 | goto retry; | |
2824 | ||
1da177e4 LT |
2825 | mark_inode_dirty(inode); |
2826 | return 0; | |
1da177e4 LT |
2827 | fail: |
2828 | return err; | |
2829 | } | |
2830 | ||
0adb25d2 KK |
2831 | int page_symlink(struct inode *inode, const char *symname, int len) |
2832 | { | |
2833 | return __page_symlink(inode, symname, len, | |
2834 | mapping_gfp_mask(inode->i_mapping)); | |
2835 | } | |
2836 | ||
92e1d5be | 2837 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2838 | .readlink = generic_readlink, |
2839 | .follow_link = page_follow_link_light, | |
2840 | .put_link = page_put_link, | |
2841 | }; | |
2842 | ||
2d8f3038 | 2843 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2844 | EXPORT_SYMBOL(follow_down); |
2845 | EXPORT_SYMBOL(follow_up); | |
2846 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2847 | EXPORT_SYMBOL(getname); | |
2848 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2849 | EXPORT_SYMBOL(lookup_one_len); |
2850 | EXPORT_SYMBOL(page_follow_link_light); | |
2851 | EXPORT_SYMBOL(page_put_link); | |
2852 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2853 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2854 | EXPORT_SYMBOL(page_symlink); |
2855 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2856 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2857 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2858 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2859 | EXPORT_SYMBOL(inode_permission); |
e4543edd | 2860 | EXPORT_SYMBOL(vfs_permission); |
8c744fb8 | 2861 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2862 | EXPORT_SYMBOL(unlock_rename); |
2863 | EXPORT_SYMBOL(vfs_create); | |
2864 | EXPORT_SYMBOL(vfs_follow_link); | |
2865 | EXPORT_SYMBOL(vfs_link); | |
2866 | EXPORT_SYMBOL(vfs_mkdir); | |
2867 | EXPORT_SYMBOL(vfs_mknod); | |
2868 | EXPORT_SYMBOL(generic_permission); | |
2869 | EXPORT_SYMBOL(vfs_readlink); | |
2870 | EXPORT_SYMBOL(vfs_rename); | |
2871 | EXPORT_SYMBOL(vfs_rmdir); | |
2872 | EXPORT_SYMBOL(vfs_symlink); | |
2873 | EXPORT_SYMBOL(vfs_unlink); | |
2874 | EXPORT_SYMBOL(dentry_unhash); | |
2875 | EXPORT_SYMBOL(generic_readlink); |