]>
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> |
5ad4e53b | 35 | #include <linux/fs_struct.h> |
1da177e4 LT |
36 | #include <asm/uaccess.h> |
37 | ||
38 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
39 | ||
40 | /* [Feb-1997 T. Schoebel-Theuer] | |
41 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
42 | * were necessary because of omirr. The reason is that omirr needs | |
43 | * to know the _real_ pathname, not the user-supplied one, in case | |
44 | * of symlinks (and also when transname replacements occur). | |
45 | * | |
46 | * The new code replaces the old recursive symlink resolution with | |
47 | * an iterative one (in case of non-nested symlink chains). It does | |
48 | * this with calls to <fs>_follow_link(). | |
49 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
50 | * replaced with a single function lookup_dentry() that can handle all | |
51 | * the special cases of the former code. | |
52 | * | |
53 | * With the new dcache, the pathname is stored at each inode, at least as | |
54 | * long as the refcount of the inode is positive. As a side effect, the | |
55 | * size of the dcache depends on the inode cache and thus is dynamic. | |
56 | * | |
57 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
58 | * resolution to correspond with current state of the code. | |
59 | * | |
60 | * Note that the symlink resolution is not *completely* iterative. | |
61 | * There is still a significant amount of tail- and mid- recursion in | |
62 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
63 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
64 | * may return different results than <fs>_follow_link(). Many virtual | |
65 | * filesystems (including /proc) exhibit this behavior. | |
66 | */ | |
67 | ||
68 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
69 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
70 | * and the name already exists in form of a symlink, try to create the new | |
71 | * name indicated by the symlink. The old code always complained that the | |
72 | * name already exists, due to not following the symlink even if its target | |
73 | * is nonexistent. The new semantics affects also mknod() and link() when | |
74 | * the name is a symlink pointing to a non-existant name. | |
75 | * | |
76 | * I don't know which semantics is the right one, since I have no access | |
77 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
78 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
79 | * "old" one. Personally, I think the new semantics is much more logical. | |
80 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
81 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
82 | * and in the old Linux semantics. | |
83 | */ | |
84 | ||
85 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
86 | * semantics. See the comments in "open_namei" and "do_link" below. | |
87 | * | |
88 | * [10-Sep-98 Alan Modra] Another symlink change. | |
89 | */ | |
90 | ||
91 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
92 | * inside the path - always follow. | |
93 | * in the last component in creation/removal/renaming - never follow. | |
94 | * if LOOKUP_FOLLOW passed - follow. | |
95 | * if the pathname has trailing slashes - follow. | |
96 | * otherwise - don't follow. | |
97 | * (applied in that order). | |
98 | * | |
99 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
100 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
101 | * During the 2.4 we need to fix the userland stuff depending on it - | |
102 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
103 | * XEmacs seems to be relying on it... | |
104 | */ | |
105 | /* | |
106 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 107 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
108 | * any extra contention... |
109 | */ | |
110 | ||
111 | /* In order to reduce some races, while at the same time doing additional | |
112 | * checking and hopefully speeding things up, we copy filenames to the | |
113 | * kernel data space before using them.. | |
114 | * | |
115 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
116 | * PATH_MAX includes the nul terminator --RR. | |
117 | */ | |
858119e1 | 118 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
119 | { |
120 | int retval; | |
121 | unsigned long len = PATH_MAX; | |
122 | ||
123 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
124 | if ((unsigned long) filename >= TASK_SIZE) | |
125 | return -EFAULT; | |
126 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
127 | len = TASK_SIZE - (unsigned long) filename; | |
128 | } | |
129 | ||
130 | retval = strncpy_from_user(page, filename, len); | |
131 | if (retval > 0) { | |
132 | if (retval < len) | |
133 | return 0; | |
134 | return -ENAMETOOLONG; | |
135 | } else if (!retval) | |
136 | retval = -ENOENT; | |
137 | return retval; | |
138 | } | |
139 | ||
140 | char * getname(const char __user * filename) | |
141 | { | |
142 | char *tmp, *result; | |
143 | ||
144 | result = ERR_PTR(-ENOMEM); | |
145 | tmp = __getname(); | |
146 | if (tmp) { | |
147 | int retval = do_getname(filename, tmp); | |
148 | ||
149 | result = tmp; | |
150 | if (retval < 0) { | |
151 | __putname(tmp); | |
152 | result = ERR_PTR(retval); | |
153 | } | |
154 | } | |
155 | audit_getname(result); | |
156 | return result; | |
157 | } | |
158 | ||
159 | #ifdef CONFIG_AUDITSYSCALL | |
160 | void putname(const char *name) | |
161 | { | |
5ac3a9c2 | 162 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
163 | audit_putname(name); |
164 | else | |
165 | __putname(name); | |
166 | } | |
167 | EXPORT_SYMBOL(putname); | |
168 | #endif | |
169 | ||
5909ccaa LT |
170 | /* |
171 | * This does basic POSIX ACL permission checking | |
1da177e4 | 172 | */ |
5909ccaa | 173 | static int acl_permission_check(struct inode *inode, int mask, |
1da177e4 LT |
174 | int (*check_acl)(struct inode *inode, int mask)) |
175 | { | |
176 | umode_t mode = inode->i_mode; | |
177 | ||
e6305c43 AV |
178 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
179 | ||
da9592ed | 180 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
181 | mode >>= 6; |
182 | else { | |
183 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
184 | int error = check_acl(inode, mask); | |
5909ccaa | 185 | if (error != -EAGAIN) |
1da177e4 LT |
186 | return error; |
187 | } | |
188 | ||
189 | if (in_group_p(inode->i_gid)) | |
190 | mode >>= 3; | |
191 | } | |
192 | ||
193 | /* | |
194 | * If the DACs are ok we don't need any capability check. | |
195 | */ | |
e6305c43 | 196 | if ((mask & ~mode) == 0) |
1da177e4 | 197 | return 0; |
5909ccaa LT |
198 | return -EACCES; |
199 | } | |
200 | ||
201 | /** | |
202 | * generic_permission - check for access rights on a Posix-like filesystem | |
203 | * @inode: inode to check access rights for | |
204 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
205 | * @check_acl: optional callback to check for Posix ACLs | |
206 | * | |
207 | * Used to check for read/write/execute permissions on a file. | |
208 | * We use "fsuid" for this, letting us set arbitrary permissions | |
209 | * for filesystem access without changing the "normal" uids which | |
210 | * are used for other things.. | |
211 | */ | |
212 | int generic_permission(struct inode *inode, int mask, | |
213 | int (*check_acl)(struct inode *inode, int mask)) | |
214 | { | |
215 | int ret; | |
216 | ||
217 | /* | |
218 | * Do the basic POSIX ACL permission checks. | |
219 | */ | |
220 | ret = acl_permission_check(inode, mask, check_acl); | |
221 | if (ret != -EACCES) | |
222 | return ret; | |
1da177e4 | 223 | |
1da177e4 LT |
224 | /* |
225 | * Read/write DACs are always overridable. | |
226 | * Executable DACs are overridable if at least one exec bit is set. | |
227 | */ | |
f696a365 | 228 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
229 | if (capable(CAP_DAC_OVERRIDE)) |
230 | return 0; | |
231 | ||
232 | /* | |
233 | * Searching includes executable on directories, else just read. | |
234 | */ | |
235 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
236 | if (capable(CAP_DAC_READ_SEARCH)) | |
237 | return 0; | |
238 | ||
239 | return -EACCES; | |
240 | } | |
241 | ||
cb23beb5 CH |
242 | /** |
243 | * inode_permission - check for access rights to a given inode | |
244 | * @inode: inode to check permission on | |
245 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
246 | * | |
247 | * Used to check for read/write/execute permissions on an inode. | |
248 | * We use "fsuid" for this, letting us set arbitrary permissions | |
249 | * for filesystem access without changing the "normal" uids which | |
250 | * are used for other things. | |
251 | */ | |
f419a2e3 | 252 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 253 | { |
e6305c43 | 254 | int retval; |
1da177e4 LT |
255 | |
256 | if (mask & MAY_WRITE) { | |
22590e41 | 257 | umode_t mode = inode->i_mode; |
1da177e4 LT |
258 | |
259 | /* | |
260 | * Nobody gets write access to a read-only fs. | |
261 | */ | |
262 | if (IS_RDONLY(inode) && | |
263 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
264 | return -EROFS; | |
265 | ||
266 | /* | |
267 | * Nobody gets write access to an immutable file. | |
268 | */ | |
269 | if (IS_IMMUTABLE(inode)) | |
270 | return -EACCES; | |
271 | } | |
272 | ||
acfa4380 | 273 | if (inode->i_op->permission) |
b77b0646 | 274 | retval = inode->i_op->permission(inode, mask); |
f696a365 | 275 | else |
5909ccaa | 276 | retval = generic_permission(inode, mask, inode->i_op->check_acl); |
f696a365 | 277 | |
1da177e4 LT |
278 | if (retval) |
279 | return retval; | |
280 | ||
08ce5f16 SH |
281 | retval = devcgroup_inode_permission(inode, mask); |
282 | if (retval) | |
283 | return retval; | |
284 | ||
e6305c43 | 285 | return security_inode_permission(inode, |
f418b006 | 286 | mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND)); |
1da177e4 LT |
287 | } |
288 | ||
8c744fb8 CH |
289 | /** |
290 | * file_permission - check for additional access rights to a given file | |
291 | * @file: file to check access rights for | |
292 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
293 | * | |
294 | * Used to check for read/write/execute permissions on an already opened | |
295 | * file. | |
296 | * | |
297 | * Note: | |
298 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 299 | * be done using inode_permission(). |
8c744fb8 CH |
300 | */ |
301 | int file_permission(struct file *file, int mask) | |
302 | { | |
f419a2e3 | 303 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
304 | } |
305 | ||
1da177e4 LT |
306 | /* |
307 | * get_write_access() gets write permission for a file. | |
308 | * put_write_access() releases this write permission. | |
309 | * This is used for regular files. | |
310 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
311 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
312 | * can have the following values: | |
313 | * 0: no writers, no VM_DENYWRITE mappings | |
314 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
315 | * > 0: (i_writecount) users are writing to the file. | |
316 | * | |
317 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
318 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
319 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
320 | * to do the change if sign is wrong. Exclusion between them is provided by | |
321 | * the inode->i_lock spinlock. | |
322 | */ | |
323 | ||
324 | int get_write_access(struct inode * inode) | |
325 | { | |
326 | spin_lock(&inode->i_lock); | |
327 | if (atomic_read(&inode->i_writecount) < 0) { | |
328 | spin_unlock(&inode->i_lock); | |
329 | return -ETXTBSY; | |
330 | } | |
331 | atomic_inc(&inode->i_writecount); | |
332 | spin_unlock(&inode->i_lock); | |
333 | ||
334 | return 0; | |
335 | } | |
336 | ||
337 | int deny_write_access(struct file * file) | |
338 | { | |
0f7fc9e4 | 339 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
340 | |
341 | spin_lock(&inode->i_lock); | |
342 | if (atomic_read(&inode->i_writecount) > 0) { | |
343 | spin_unlock(&inode->i_lock); | |
344 | return -ETXTBSY; | |
345 | } | |
346 | atomic_dec(&inode->i_writecount); | |
347 | spin_unlock(&inode->i_lock); | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
5dd784d0 JB |
352 | /** |
353 | * path_get - get a reference to a path | |
354 | * @path: path to get the reference to | |
355 | * | |
356 | * Given a path increment the reference count to the dentry and the vfsmount. | |
357 | */ | |
358 | void path_get(struct path *path) | |
359 | { | |
360 | mntget(path->mnt); | |
361 | dget(path->dentry); | |
362 | } | |
363 | EXPORT_SYMBOL(path_get); | |
364 | ||
1d957f9b JB |
365 | /** |
366 | * path_put - put a reference to a path | |
367 | * @path: path to put the reference to | |
368 | * | |
369 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
370 | */ | |
371 | void path_put(struct path *path) | |
1da177e4 | 372 | { |
1d957f9b JB |
373 | dput(path->dentry); |
374 | mntput(path->mnt); | |
1da177e4 | 375 | } |
1d957f9b | 376 | EXPORT_SYMBOL(path_put); |
1da177e4 | 377 | |
834f2a4a TM |
378 | /** |
379 | * release_open_intent - free up open intent resources | |
380 | * @nd: pointer to nameidata | |
381 | */ | |
382 | void release_open_intent(struct nameidata *nd) | |
383 | { | |
0f7fc9e4 | 384 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
385 | put_filp(nd->intent.open.file); |
386 | else | |
387 | fput(nd->intent.open.file); | |
388 | } | |
389 | ||
bcdc5e01 IK |
390 | static inline struct dentry * |
391 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
392 | { | |
393 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
394 | if (unlikely(status <= 0)) { | |
395 | /* | |
396 | * The dentry failed validation. | |
397 | * If d_revalidate returned 0 attempt to invalidate | |
398 | * the dentry otherwise d_revalidate is asking us | |
399 | * to return a fail status. | |
400 | */ | |
401 | if (!status) { | |
402 | if (!d_invalidate(dentry)) { | |
403 | dput(dentry); | |
404 | dentry = NULL; | |
405 | } | |
406 | } else { | |
407 | dput(dentry); | |
408 | dentry = ERR_PTR(status); | |
409 | } | |
410 | } | |
411 | return dentry; | |
412 | } | |
413 | ||
1da177e4 LT |
414 | /* |
415 | * Internal lookup() using the new generic dcache. | |
416 | * SMP-safe | |
417 | */ | |
418 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
419 | { | |
420 | struct dentry * dentry = __d_lookup(parent, name); | |
421 | ||
422 | /* lockess __d_lookup may fail due to concurrent d_move() | |
423 | * in some unrelated directory, so try with d_lookup | |
424 | */ | |
425 | if (!dentry) | |
426 | dentry = d_lookup(parent, name); | |
427 | ||
bcdc5e01 IK |
428 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
429 | dentry = do_revalidate(dentry, nd); | |
430 | ||
1da177e4 LT |
431 | return dentry; |
432 | } | |
433 | ||
434 | /* | |
435 | * Short-cut version of permission(), for calling by | |
436 | * path_walk(), when dcache lock is held. Combines parts | |
437 | * of permission() and generic_permission(), and tests ONLY for | |
438 | * MAY_EXEC permission. | |
439 | * | |
440 | * If appropriate, check DAC only. If not appropriate, or | |
441 | * short-cut DAC fails, then call permission() to do more | |
442 | * complete permission check. | |
443 | */ | |
672b16b2 | 444 | static int exec_permission_lite(struct inode *inode) |
1da177e4 | 445 | { |
5909ccaa | 446 | int ret; |
1da177e4 | 447 | |
cb9179ea | 448 | if (inode->i_op->permission) { |
5909ccaa | 449 | ret = inode->i_op->permission(inode, MAY_EXEC); |
cb9179ea LT |
450 | if (!ret) |
451 | goto ok; | |
452 | return ret; | |
453 | } | |
5909ccaa LT |
454 | ret = acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl); |
455 | if (!ret) | |
1da177e4 LT |
456 | goto ok; |
457 | ||
f1ac9f6b | 458 | if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
459 | goto ok; |
460 | ||
5909ccaa | 461 | return ret; |
1da177e4 | 462 | ok: |
b77b0646 | 463 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
464 | } |
465 | ||
466 | /* | |
467 | * This is called when everything else fails, and we actually have | |
468 | * to go to the low-level filesystem to find out what we should do.. | |
469 | * | |
470 | * We get the directory semaphore, and after getting that we also | |
471 | * make sure that nobody added the entry to the dcache in the meantime.. | |
472 | * SMP-safe | |
473 | */ | |
474 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
475 | { | |
476 | struct dentry * result; | |
477 | struct inode *dir = parent->d_inode; | |
478 | ||
1b1dcc1b | 479 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
480 | /* |
481 | * First re-do the cached lookup just in case it was created | |
482 | * while we waited for the directory semaphore.. | |
483 | * | |
484 | * FIXME! This could use version numbering or similar to | |
485 | * avoid unnecessary cache lookups. | |
486 | * | |
487 | * The "dcache_lock" is purely to protect the RCU list walker | |
488 | * from concurrent renames at this point (we mustn't get false | |
489 | * negatives from the RCU list walk here, unlike the optimistic | |
490 | * fast walk). | |
491 | * | |
492 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
493 | */ | |
494 | result = d_lookup(parent, name); | |
495 | if (!result) { | |
d70b67c8 MS |
496 | struct dentry *dentry; |
497 | ||
498 | /* Don't create child dentry for a dead directory. */ | |
499 | result = ERR_PTR(-ENOENT); | |
500 | if (IS_DEADDIR(dir)) | |
501 | goto out_unlock; | |
502 | ||
503 | dentry = d_alloc(parent, name); | |
1da177e4 LT |
504 | result = ERR_PTR(-ENOMEM); |
505 | if (dentry) { | |
506 | result = dir->i_op->lookup(dir, dentry, nd); | |
507 | if (result) | |
508 | dput(dentry); | |
509 | else | |
510 | result = dentry; | |
511 | } | |
d70b67c8 | 512 | out_unlock: |
1b1dcc1b | 513 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
514 | return result; |
515 | } | |
516 | ||
517 | /* | |
518 | * Uhhuh! Nasty case: the cache was re-populated while | |
519 | * we waited on the semaphore. Need to revalidate. | |
520 | */ | |
1b1dcc1b | 521 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 522 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
523 | result = do_revalidate(result, nd); |
524 | if (!result) | |
1da177e4 | 525 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
526 | } |
527 | return result; | |
528 | } | |
529 | ||
2a737871 AV |
530 | static __always_inline void set_root(struct nameidata *nd) |
531 | { | |
532 | if (!nd->root.mnt) { | |
533 | struct fs_struct *fs = current->fs; | |
534 | read_lock(&fs->lock); | |
535 | nd->root = fs->root; | |
536 | path_get(&nd->root); | |
537 | read_unlock(&fs->lock); | |
538 | } | |
539 | } | |
540 | ||
6de88d72 AV |
541 | static int link_path_walk(const char *, struct nameidata *); |
542 | ||
f1662356 | 543 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
544 | { |
545 | int res = 0; | |
546 | char *name; | |
547 | if (IS_ERR(link)) | |
548 | goto fail; | |
549 | ||
550 | if (*link == '/') { | |
2a737871 | 551 | set_root(nd); |
1d957f9b | 552 | path_put(&nd->path); |
2a737871 AV |
553 | nd->path = nd->root; |
554 | path_get(&nd->root); | |
1da177e4 | 555 | } |
b4091d5f | 556 | |
1da177e4 | 557 | res = link_path_walk(link, nd); |
1da177e4 LT |
558 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
559 | return res; | |
560 | /* | |
561 | * If it is an iterative symlinks resolution in open_namei() we | |
562 | * have to copy the last component. And all that crap because of | |
563 | * bloody create() on broken symlinks. Furrfu... | |
564 | */ | |
565 | name = __getname(); | |
566 | if (unlikely(!name)) { | |
1d957f9b | 567 | path_put(&nd->path); |
1da177e4 LT |
568 | return -ENOMEM; |
569 | } | |
570 | strcpy(name, nd->last.name); | |
571 | nd->last.name = name; | |
572 | return 0; | |
573 | fail: | |
1d957f9b | 574 | path_put(&nd->path); |
1da177e4 LT |
575 | return PTR_ERR(link); |
576 | } | |
577 | ||
1d957f9b | 578 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
579 | { |
580 | dput(path->dentry); | |
4ac91378 | 581 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
582 | mntput(path->mnt); |
583 | } | |
584 | ||
585 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
586 | { | |
4ac91378 JB |
587 | dput(nd->path.dentry); |
588 | if (nd->path.mnt != path->mnt) | |
589 | mntput(nd->path.mnt); | |
590 | nd->path.mnt = path->mnt; | |
591 | nd->path.dentry = path->dentry; | |
051d3812 IK |
592 | } |
593 | ||
f1662356 | 594 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
595 | { |
596 | int error; | |
cc314eef | 597 | void *cookie; |
cd4e91d3 | 598 | struct dentry *dentry = path->dentry; |
1da177e4 | 599 | |
d671a1cb | 600 | touch_atime(path->mnt, dentry); |
1da177e4 | 601 | nd_set_link(nd, NULL); |
cd4e91d3 | 602 | |
4ac91378 | 603 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
604 | path_to_nameidata(path, nd); |
605 | dget(dentry); | |
606 | } | |
607 | mntget(path->mnt); | |
cc314eef LT |
608 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
609 | error = PTR_ERR(cookie); | |
610 | if (!IS_ERR(cookie)) { | |
1da177e4 | 611 | char *s = nd_get_link(nd); |
cc314eef | 612 | error = 0; |
1da177e4 LT |
613 | if (s) |
614 | error = __vfs_follow_link(nd, s); | |
615 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 616 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 617 | } |
1da177e4 LT |
618 | return error; |
619 | } | |
620 | ||
621 | /* | |
622 | * This limits recursive symlink follows to 8, while | |
623 | * limiting consecutive symlinks to 40. | |
624 | * | |
625 | * Without that kind of total limit, nasty chains of consecutive | |
626 | * symlinks can cause almost arbitrarily long lookups. | |
627 | */ | |
90ebe565 | 628 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
629 | { |
630 | int err = -ELOOP; | |
631 | if (current->link_count >= MAX_NESTED_LINKS) | |
632 | goto loop; | |
633 | if (current->total_link_count >= 40) | |
634 | goto loop; | |
635 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
636 | cond_resched(); | |
90ebe565 | 637 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
638 | if (err) |
639 | goto loop; | |
640 | current->link_count++; | |
641 | current->total_link_count++; | |
642 | nd->depth++; | |
cd4e91d3 | 643 | err = __do_follow_link(path, nd); |
258fa999 | 644 | path_put(path); |
839d9f93 AV |
645 | current->link_count--; |
646 | nd->depth--; | |
1da177e4 LT |
647 | return err; |
648 | loop: | |
1d957f9b JB |
649 | path_put_conditional(path, nd); |
650 | path_put(&nd->path); | |
1da177e4 LT |
651 | return err; |
652 | } | |
653 | ||
bab77ebf | 654 | int follow_up(struct path *path) |
1da177e4 LT |
655 | { |
656 | struct vfsmount *parent; | |
657 | struct dentry *mountpoint; | |
658 | spin_lock(&vfsmount_lock); | |
bab77ebf AV |
659 | parent = path->mnt->mnt_parent; |
660 | if (parent == path->mnt) { | |
1da177e4 LT |
661 | spin_unlock(&vfsmount_lock); |
662 | return 0; | |
663 | } | |
664 | mntget(parent); | |
bab77ebf | 665 | mountpoint = dget(path->mnt->mnt_mountpoint); |
1da177e4 | 666 | spin_unlock(&vfsmount_lock); |
bab77ebf AV |
667 | dput(path->dentry); |
668 | path->dentry = mountpoint; | |
669 | mntput(path->mnt); | |
670 | path->mnt = parent; | |
1da177e4 LT |
671 | return 1; |
672 | } | |
673 | ||
674 | /* no need for dcache_lock, as serialization is taken care in | |
675 | * namespace.c | |
676 | */ | |
463ffb2e AV |
677 | static int __follow_mount(struct path *path) |
678 | { | |
679 | int res = 0; | |
680 | while (d_mountpoint(path->dentry)) { | |
1c755af4 | 681 | struct vfsmount *mounted = lookup_mnt(path); |
463ffb2e AV |
682 | if (!mounted) |
683 | break; | |
684 | dput(path->dentry); | |
685 | if (res) | |
686 | mntput(path->mnt); | |
687 | path->mnt = mounted; | |
688 | path->dentry = dget(mounted->mnt_root); | |
689 | res = 1; | |
690 | } | |
691 | return res; | |
692 | } | |
693 | ||
79ed0226 | 694 | static void follow_mount(struct path *path) |
1da177e4 | 695 | { |
79ed0226 | 696 | while (d_mountpoint(path->dentry)) { |
1c755af4 | 697 | struct vfsmount *mounted = lookup_mnt(path); |
1da177e4 LT |
698 | if (!mounted) |
699 | break; | |
79ed0226 AV |
700 | dput(path->dentry); |
701 | mntput(path->mnt); | |
702 | path->mnt = mounted; | |
703 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 | 704 | } |
1da177e4 LT |
705 | } |
706 | ||
707 | /* no need for dcache_lock, as serialization is taken care in | |
708 | * namespace.c | |
709 | */ | |
9393bd07 | 710 | int follow_down(struct path *path) |
1da177e4 LT |
711 | { |
712 | struct vfsmount *mounted; | |
713 | ||
1c755af4 | 714 | mounted = lookup_mnt(path); |
1da177e4 | 715 | if (mounted) { |
9393bd07 AV |
716 | dput(path->dentry); |
717 | mntput(path->mnt); | |
718 | path->mnt = mounted; | |
719 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
720 | return 1; |
721 | } | |
722 | return 0; | |
723 | } | |
724 | ||
f1662356 | 725 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 726 | { |
2a737871 | 727 | set_root(nd); |
e518ddb7 | 728 | |
1da177e4 LT |
729 | while(1) { |
730 | struct vfsmount *parent; | |
4ac91378 | 731 | struct dentry *old = nd->path.dentry; |
1da177e4 | 732 | |
2a737871 AV |
733 | if (nd->path.dentry == nd->root.dentry && |
734 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
735 | break; |
736 | } | |
1da177e4 | 737 | spin_lock(&dcache_lock); |
4ac91378 JB |
738 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
739 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
740 | spin_unlock(&dcache_lock); |
741 | dput(old); | |
742 | break; | |
743 | } | |
744 | spin_unlock(&dcache_lock); | |
745 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
746 | parent = nd->path.mnt->mnt_parent; |
747 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
748 | spin_unlock(&vfsmount_lock); |
749 | break; | |
750 | } | |
751 | mntget(parent); | |
4ac91378 | 752 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
753 | spin_unlock(&vfsmount_lock); |
754 | dput(old); | |
4ac91378 JB |
755 | mntput(nd->path.mnt); |
756 | nd->path.mnt = parent; | |
1da177e4 | 757 | } |
79ed0226 | 758 | follow_mount(&nd->path); |
1da177e4 LT |
759 | } |
760 | ||
1da177e4 LT |
761 | /* |
762 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
763 | * small and for now I'd prefer to have fast path as straight as possible. | |
764 | * It _is_ time-critical. | |
765 | */ | |
766 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
767 | struct path *path) | |
768 | { | |
4ac91378 | 769 | struct vfsmount *mnt = nd->path.mnt; |
3cac260a AV |
770 | struct dentry *dentry; |
771 | /* | |
772 | * See if the low-level filesystem might want | |
773 | * to use its own hash.. | |
774 | */ | |
775 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { | |
776 | int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, name); | |
777 | if (err < 0) | |
778 | return err; | |
779 | } | |
1da177e4 | 780 | |
3cac260a | 781 | dentry = __d_lookup(nd->path.dentry, name); |
1da177e4 LT |
782 | if (!dentry) |
783 | goto need_lookup; | |
784 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
785 | goto need_revalidate; | |
786 | done: | |
787 | path->mnt = mnt; | |
788 | path->dentry = dentry; | |
634ee701 | 789 | __follow_mount(path); |
1da177e4 LT |
790 | return 0; |
791 | ||
792 | need_lookup: | |
4ac91378 | 793 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
794 | if (IS_ERR(dentry)) |
795 | goto fail; | |
796 | goto done; | |
797 | ||
798 | need_revalidate: | |
bcdc5e01 IK |
799 | dentry = do_revalidate(dentry, nd); |
800 | if (!dentry) | |
801 | goto need_lookup; | |
802 | if (IS_ERR(dentry)) | |
803 | goto fail; | |
804 | goto done; | |
1da177e4 LT |
805 | |
806 | fail: | |
807 | return PTR_ERR(dentry); | |
808 | } | |
809 | ||
810 | /* | |
811 | * Name resolution. | |
ea3834d9 PM |
812 | * This is the basic name resolution function, turning a pathname into |
813 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 814 | * |
ea3834d9 PM |
815 | * Returns 0 and nd will have valid dentry and mnt on success. |
816 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 817 | */ |
6de88d72 | 818 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
819 | { |
820 | struct path next; | |
821 | struct inode *inode; | |
822 | int err; | |
823 | unsigned int lookup_flags = nd->flags; | |
824 | ||
825 | while (*name=='/') | |
826 | name++; | |
827 | if (!*name) | |
828 | goto return_reval; | |
829 | ||
4ac91378 | 830 | inode = nd->path.dentry->d_inode; |
1da177e4 | 831 | if (nd->depth) |
f55eab82 | 832 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
833 | |
834 | /* At this point we know we have a real path component. */ | |
835 | for(;;) { | |
836 | unsigned long hash; | |
837 | struct qstr this; | |
838 | unsigned int c; | |
839 | ||
cdce5d6b | 840 | nd->flags |= LOOKUP_CONTINUE; |
672b16b2 | 841 | err = exec_permission_lite(inode); |
1da177e4 LT |
842 | if (err) |
843 | break; | |
844 | ||
845 | this.name = name; | |
846 | c = *(const unsigned char *)name; | |
847 | ||
848 | hash = init_name_hash(); | |
849 | do { | |
850 | name++; | |
851 | hash = partial_name_hash(c, hash); | |
852 | c = *(const unsigned char *)name; | |
853 | } while (c && (c != '/')); | |
854 | this.len = name - (const char *) this.name; | |
855 | this.hash = end_name_hash(hash); | |
856 | ||
857 | /* remove trailing slashes? */ | |
858 | if (!c) | |
859 | goto last_component; | |
860 | while (*++name == '/'); | |
861 | if (!*name) | |
862 | goto last_with_slashes; | |
863 | ||
864 | /* | |
865 | * "." and ".." are special - ".." especially so because it has | |
866 | * to be able to know about the current root directory and | |
867 | * parent relationships. | |
868 | */ | |
869 | if (this.name[0] == '.') switch (this.len) { | |
870 | default: | |
871 | break; | |
872 | case 2: | |
873 | if (this.name[1] != '.') | |
874 | break; | |
58c465eb | 875 | follow_dotdot(nd); |
4ac91378 | 876 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
877 | /* fallthrough */ |
878 | case 1: | |
879 | continue; | |
880 | } | |
1da177e4 LT |
881 | /* This does the actual lookups.. */ |
882 | err = do_lookup(nd, &this, &next); | |
883 | if (err) | |
884 | break; | |
1da177e4 LT |
885 | |
886 | err = -ENOENT; | |
887 | inode = next.dentry->d_inode; | |
888 | if (!inode) | |
889 | goto out_dput; | |
1da177e4 LT |
890 | |
891 | if (inode->i_op->follow_link) { | |
90ebe565 | 892 | err = do_follow_link(&next, nd); |
1da177e4 LT |
893 | if (err) |
894 | goto return_err; | |
895 | err = -ENOENT; | |
4ac91378 | 896 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
897 | if (!inode) |
898 | break; | |
09dd17d3 MS |
899 | } else |
900 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
901 | err = -ENOTDIR; |
902 | if (!inode->i_op->lookup) | |
903 | break; | |
904 | continue; | |
905 | /* here ends the main loop */ | |
906 | ||
907 | last_with_slashes: | |
908 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
909 | last_component: | |
f55eab82 TM |
910 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
911 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
912 | if (lookup_flags & LOOKUP_PARENT) |
913 | goto lookup_parent; | |
914 | if (this.name[0] == '.') switch (this.len) { | |
915 | default: | |
916 | break; | |
917 | case 2: | |
918 | if (this.name[1] != '.') | |
919 | break; | |
58c465eb | 920 | follow_dotdot(nd); |
4ac91378 | 921 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
922 | /* fallthrough */ |
923 | case 1: | |
924 | goto return_reval; | |
925 | } | |
1da177e4 LT |
926 | err = do_lookup(nd, &this, &next); |
927 | if (err) | |
928 | break; | |
1da177e4 LT |
929 | inode = next.dentry->d_inode; |
930 | if ((lookup_flags & LOOKUP_FOLLOW) | |
acfa4380 | 931 | && inode && inode->i_op->follow_link) { |
90ebe565 | 932 | err = do_follow_link(&next, nd); |
1da177e4 LT |
933 | if (err) |
934 | goto return_err; | |
4ac91378 | 935 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
936 | } else |
937 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
938 | err = -ENOENT; |
939 | if (!inode) | |
940 | break; | |
941 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
942 | err = -ENOTDIR; | |
acfa4380 | 943 | if (!inode->i_op->lookup) |
1da177e4 LT |
944 | break; |
945 | } | |
946 | goto return_base; | |
947 | lookup_parent: | |
948 | nd->last = this; | |
949 | nd->last_type = LAST_NORM; | |
950 | if (this.name[0] != '.') | |
951 | goto return_base; | |
952 | if (this.len == 1) | |
953 | nd->last_type = LAST_DOT; | |
954 | else if (this.len == 2 && this.name[1] == '.') | |
955 | nd->last_type = LAST_DOTDOT; | |
956 | else | |
957 | goto return_base; | |
958 | return_reval: | |
959 | /* | |
960 | * We bypassed the ordinary revalidation routines. | |
961 | * We may need to check the cached dentry for staleness. | |
962 | */ | |
4ac91378 JB |
963 | if (nd->path.dentry && nd->path.dentry->d_sb && |
964 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
965 | err = -ESTALE; |
966 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
967 | if (!nd->path.dentry->d_op->d_revalidate( |
968 | nd->path.dentry, nd)) | |
1da177e4 LT |
969 | break; |
970 | } | |
971 | return_base: | |
972 | return 0; | |
973 | out_dput: | |
1d957f9b | 974 | path_put_conditional(&next, nd); |
1da177e4 LT |
975 | break; |
976 | } | |
1d957f9b | 977 | path_put(&nd->path); |
1da177e4 LT |
978 | return_err: |
979 | return err; | |
980 | } | |
981 | ||
fc9b52cd | 982 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 983 | { |
6de88d72 AV |
984 | struct path save = nd->path; |
985 | int result; | |
986 | ||
1da177e4 | 987 | current->total_link_count = 0; |
6de88d72 AV |
988 | |
989 | /* make sure the stuff we saved doesn't go away */ | |
990 | path_get(&save); | |
991 | ||
992 | result = link_path_walk(name, nd); | |
993 | if (result == -ESTALE) { | |
994 | /* nd->path had been dropped */ | |
995 | current->total_link_count = 0; | |
996 | nd->path = save; | |
997 | path_get(&nd->path); | |
998 | nd->flags |= LOOKUP_REVAL; | |
999 | result = link_path_walk(name, nd); | |
1000 | } | |
1001 | ||
1002 | path_put(&save); | |
1003 | ||
1004 | return result; | |
1da177e4 LT |
1005 | } |
1006 | ||
9b4a9b14 | 1007 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1008 | { |
ea3834d9 | 1009 | int retval = 0; |
170aa3d0 UD |
1010 | int fput_needed; |
1011 | struct file *file; | |
1da177e4 LT |
1012 | |
1013 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1014 | nd->flags = flags; | |
1015 | nd->depth = 0; | |
2a737871 | 1016 | nd->root.mnt = NULL; |
1da177e4 | 1017 | |
1da177e4 | 1018 | if (*name=='/') { |
2a737871 AV |
1019 | set_root(nd); |
1020 | nd->path = nd->root; | |
1021 | path_get(&nd->root); | |
5590ff0d | 1022 | } else if (dfd == AT_FDCWD) { |
2a737871 | 1023 | struct fs_struct *fs = current->fs; |
e518ddb7 | 1024 | read_lock(&fs->lock); |
6ac08c39 JB |
1025 | nd->path = fs->pwd; |
1026 | path_get(&fs->pwd); | |
e518ddb7 | 1027 | read_unlock(&fs->lock); |
5590ff0d | 1028 | } else { |
5590ff0d UD |
1029 | struct dentry *dentry; |
1030 | ||
1031 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1032 | retval = -EBADF; |
1033 | if (!file) | |
6d09bb62 | 1034 | goto out_fail; |
5590ff0d | 1035 | |
0f7fc9e4 | 1036 | dentry = file->f_path.dentry; |
5590ff0d | 1037 | |
170aa3d0 UD |
1038 | retval = -ENOTDIR; |
1039 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1040 | goto fput_fail; |
5590ff0d UD |
1041 | |
1042 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1043 | if (retval) |
6d09bb62 | 1044 | goto fput_fail; |
5590ff0d | 1045 | |
5dd784d0 JB |
1046 | nd->path = file->f_path; |
1047 | path_get(&file->f_path); | |
5590ff0d UD |
1048 | |
1049 | fput_light(file, fput_needed); | |
1da177e4 | 1050 | } |
9b4a9b14 | 1051 | return 0; |
2dfdd266 | 1052 | |
9b4a9b14 AV |
1053 | fput_fail: |
1054 | fput_light(file, fput_needed); | |
1055 | out_fail: | |
1056 | return retval; | |
1057 | } | |
1058 | ||
1059 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1060 | static int do_path_lookup(int dfd, const char *name, | |
1061 | unsigned int flags, struct nameidata *nd) | |
1062 | { | |
1063 | int retval = path_init(dfd, name, flags, nd); | |
1064 | if (!retval) | |
1065 | retval = path_walk(name, nd); | |
4ac91378 JB |
1066 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1067 | nd->path.dentry->d_inode)) | |
1068 | audit_inode(name, nd->path.dentry); | |
2a737871 AV |
1069 | if (nd->root.mnt) { |
1070 | path_put(&nd->root); | |
1071 | nd->root.mnt = NULL; | |
1072 | } | |
170aa3d0 | 1073 | return retval; |
1da177e4 LT |
1074 | } |
1075 | ||
fc9b52cd | 1076 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1077 | struct nameidata *nd) |
1078 | { | |
1079 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1080 | } | |
1081 | ||
d1811465 AV |
1082 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1083 | { | |
1084 | struct nameidata nd; | |
1085 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1086 | if (!res) | |
1087 | *path = nd.path; | |
1088 | return res; | |
1089 | } | |
1090 | ||
16f18200 JJS |
1091 | /** |
1092 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1093 | * @dentry: pointer to dentry of the base directory | |
1094 | * @mnt: pointer to vfs mount of the base directory | |
1095 | * @name: pointer to file name | |
1096 | * @flags: lookup flags | |
1097 | * @nd: pointer to nameidata | |
1098 | */ | |
1099 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1100 | const char *name, unsigned int flags, | |
1101 | struct nameidata *nd) | |
1102 | { | |
1103 | int retval; | |
1104 | ||
1105 | /* same as do_path_lookup */ | |
1106 | nd->last_type = LAST_ROOT; | |
1107 | nd->flags = flags; | |
1108 | nd->depth = 0; | |
1109 | ||
c8e7f449 JB |
1110 | nd->path.dentry = dentry; |
1111 | nd->path.mnt = mnt; | |
1112 | path_get(&nd->path); | |
5b857119 AV |
1113 | nd->root = nd->path; |
1114 | path_get(&nd->root); | |
16f18200 JJS |
1115 | |
1116 | retval = path_walk(name, nd); | |
4ac91378 JB |
1117 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1118 | nd->path.dentry->d_inode)) | |
1119 | audit_inode(name, nd->path.dentry); | |
16f18200 | 1120 | |
5b857119 AV |
1121 | path_put(&nd->root); |
1122 | nd->root.mnt = NULL; | |
16f18200 | 1123 | |
2a737871 | 1124 | return retval; |
16f18200 JJS |
1125 | } |
1126 | ||
eead1911 CH |
1127 | static struct dentry *__lookup_hash(struct qstr *name, |
1128 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1129 | { |
057f6c01 | 1130 | struct dentry *dentry; |
1da177e4 LT |
1131 | struct inode *inode; |
1132 | int err; | |
1133 | ||
1134 | inode = base->d_inode; | |
1da177e4 LT |
1135 | |
1136 | /* | |
1137 | * See if the low-level filesystem might want | |
1138 | * to use its own hash.. | |
1139 | */ | |
1140 | if (base->d_op && base->d_op->d_hash) { | |
1141 | err = base->d_op->d_hash(base, name); | |
1142 | dentry = ERR_PTR(err); | |
1143 | if (err < 0) | |
1144 | goto out; | |
1145 | } | |
1146 | ||
1147 | dentry = cached_lookup(base, name, nd); | |
1148 | if (!dentry) { | |
d70b67c8 MS |
1149 | struct dentry *new; |
1150 | ||
1151 | /* Don't create child dentry for a dead directory. */ | |
1152 | dentry = ERR_PTR(-ENOENT); | |
1153 | if (IS_DEADDIR(inode)) | |
1154 | goto out; | |
1155 | ||
1156 | new = d_alloc(base, name); | |
1da177e4 LT |
1157 | dentry = ERR_PTR(-ENOMEM); |
1158 | if (!new) | |
1159 | goto out; | |
1160 | dentry = inode->i_op->lookup(inode, new, nd); | |
1161 | if (!dentry) | |
1162 | dentry = new; | |
1163 | else | |
1164 | dput(new); | |
1165 | } | |
1166 | out: | |
1167 | return dentry; | |
1168 | } | |
1169 | ||
057f6c01 JM |
1170 | /* |
1171 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1172 | * needs parent already locked. Doesn't follow mounts. | |
1173 | * SMP-safe. | |
1174 | */ | |
eead1911 | 1175 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1176 | { |
057f6c01 JM |
1177 | int err; |
1178 | ||
f419a2e3 | 1179 | err = inode_permission(nd->path.dentry->d_inode, MAY_EXEC); |
057f6c01 | 1180 | if (err) |
eead1911 | 1181 | return ERR_PTR(err); |
4ac91378 | 1182 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1183 | } |
1184 | ||
eead1911 CH |
1185 | static int __lookup_one_len(const char *name, struct qstr *this, |
1186 | struct dentry *base, int len) | |
1da177e4 LT |
1187 | { |
1188 | unsigned long hash; | |
1da177e4 LT |
1189 | unsigned int c; |
1190 | ||
057f6c01 JM |
1191 | this->name = name; |
1192 | this->len = len; | |
1da177e4 | 1193 | if (!len) |
057f6c01 | 1194 | return -EACCES; |
1da177e4 LT |
1195 | |
1196 | hash = init_name_hash(); | |
1197 | while (len--) { | |
1198 | c = *(const unsigned char *)name++; | |
1199 | if (c == '/' || c == '\0') | |
057f6c01 | 1200 | return -EACCES; |
1da177e4 LT |
1201 | hash = partial_name_hash(c, hash); |
1202 | } | |
057f6c01 JM |
1203 | this->hash = end_name_hash(hash); |
1204 | return 0; | |
1205 | } | |
1da177e4 | 1206 | |
eead1911 | 1207 | /** |
a6b91919 | 1208 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1209 | * @name: pathname component to lookup |
1210 | * @base: base directory to lookup from | |
1211 | * @len: maximum length @len should be interpreted to | |
1212 | * | |
a6b91919 RD |
1213 | * Note that this routine is purely a helper for filesystem usage and should |
1214 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1215 | * nameidata argument is passed to the filesystem methods and a filesystem |
1216 | * using this helper needs to be prepared for that. | |
1217 | */ | |
057f6c01 JM |
1218 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1219 | { | |
1220 | int err; | |
1221 | struct qstr this; | |
1222 | ||
2f9092e1 DW |
1223 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1224 | ||
057f6c01 | 1225 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1226 | if (err) |
1227 | return ERR_PTR(err); | |
1228 | ||
f419a2e3 | 1229 | err = inode_permission(base->d_inode, MAY_EXEC); |
057f6c01 JM |
1230 | if (err) |
1231 | return ERR_PTR(err); | |
49705b77 | 1232 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1233 | } |
1234 | ||
2d8f3038 AV |
1235 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1236 | struct path *path) | |
1da177e4 | 1237 | { |
2d8f3038 | 1238 | struct nameidata nd; |
1da177e4 LT |
1239 | char *tmp = getname(name); |
1240 | int err = PTR_ERR(tmp); | |
1da177e4 | 1241 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1242 | |
1243 | BUG_ON(flags & LOOKUP_PARENT); | |
1244 | ||
1245 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1246 | putname(tmp); |
2d8f3038 AV |
1247 | if (!err) |
1248 | *path = nd.path; | |
1da177e4 LT |
1249 | } |
1250 | return err; | |
1251 | } | |
1252 | ||
2ad94ae6 AV |
1253 | static int user_path_parent(int dfd, const char __user *path, |
1254 | struct nameidata *nd, char **name) | |
1255 | { | |
1256 | char *s = getname(path); | |
1257 | int error; | |
1258 | ||
1259 | if (IS_ERR(s)) | |
1260 | return PTR_ERR(s); | |
1261 | ||
1262 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1263 | if (error) | |
1264 | putname(s); | |
1265 | else | |
1266 | *name = s; | |
1267 | ||
1268 | return error; | |
1269 | } | |
1270 | ||
1da177e4 LT |
1271 | /* |
1272 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1273 | * minimal. | |
1274 | */ | |
1275 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1276 | { | |
da9592ed DH |
1277 | uid_t fsuid = current_fsuid(); |
1278 | ||
1da177e4 LT |
1279 | if (!(dir->i_mode & S_ISVTX)) |
1280 | return 0; | |
da9592ed | 1281 | if (inode->i_uid == fsuid) |
1da177e4 | 1282 | return 0; |
da9592ed | 1283 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1284 | return 0; |
1285 | return !capable(CAP_FOWNER); | |
1286 | } | |
1287 | ||
1288 | /* | |
1289 | * Check whether we can remove a link victim from directory dir, check | |
1290 | * whether the type of victim is right. | |
1291 | * 1. We can't do it if dir is read-only (done in permission()) | |
1292 | * 2. We should have write and exec permissions on dir | |
1293 | * 3. We can't remove anything from append-only dir | |
1294 | * 4. We can't do anything with immutable dir (done in permission()) | |
1295 | * 5. If the sticky bit on dir is set we should either | |
1296 | * a. be owner of dir, or | |
1297 | * b. be owner of victim, or | |
1298 | * c. have CAP_FOWNER capability | |
1299 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1300 | * links pointing to it. | |
1301 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1302 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1303 | * 9. We can't remove a root or mountpoint. | |
1304 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1305 | * nfs_async_unlink(). | |
1306 | */ | |
858119e1 | 1307 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1308 | { |
1309 | int error; | |
1310 | ||
1311 | if (!victim->d_inode) | |
1312 | return -ENOENT; | |
1313 | ||
1314 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1315 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1316 | |
f419a2e3 | 1317 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1318 | if (error) |
1319 | return error; | |
1320 | if (IS_APPEND(dir)) | |
1321 | return -EPERM; | |
1322 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1323 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1324 | return -EPERM; |
1325 | if (isdir) { | |
1326 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1327 | return -ENOTDIR; | |
1328 | if (IS_ROOT(victim)) | |
1329 | return -EBUSY; | |
1330 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1331 | return -EISDIR; | |
1332 | if (IS_DEADDIR(dir)) | |
1333 | return -ENOENT; | |
1334 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1335 | return -EBUSY; | |
1336 | return 0; | |
1337 | } | |
1338 | ||
1339 | /* Check whether we can create an object with dentry child in directory | |
1340 | * dir. | |
1341 | * 1. We can't do it if child already exists (open has special treatment for | |
1342 | * this case, but since we are inlined it's OK) | |
1343 | * 2. We can't do it if dir is read-only (done in permission()) | |
1344 | * 3. We should have write and exec permissions on dir | |
1345 | * 4. We can't do it if dir is immutable (done in permission()) | |
1346 | */ | |
a95164d9 | 1347 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1348 | { |
1349 | if (child->d_inode) | |
1350 | return -EEXIST; | |
1351 | if (IS_DEADDIR(dir)) | |
1352 | return -ENOENT; | |
f419a2e3 | 1353 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1354 | } |
1355 | ||
1356 | /* | |
1da177e4 LT |
1357 | * O_DIRECTORY translates into forcing a directory lookup. |
1358 | */ | |
1359 | static inline int lookup_flags(unsigned int f) | |
1360 | { | |
1361 | unsigned long retval = LOOKUP_FOLLOW; | |
1362 | ||
1363 | if (f & O_NOFOLLOW) | |
1364 | retval &= ~LOOKUP_FOLLOW; | |
1365 | ||
1da177e4 LT |
1366 | if (f & O_DIRECTORY) |
1367 | retval |= LOOKUP_DIRECTORY; | |
1368 | ||
1369 | return retval; | |
1370 | } | |
1371 | ||
1372 | /* | |
1373 | * p1 and p2 should be directories on the same fs. | |
1374 | */ | |
1375 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1376 | { | |
1377 | struct dentry *p; | |
1378 | ||
1379 | if (p1 == p2) { | |
f2eace23 | 1380 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1381 | return NULL; |
1382 | } | |
1383 | ||
a11f3a05 | 1384 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1385 | |
e2761a11 OH |
1386 | p = d_ancestor(p2, p1); |
1387 | if (p) { | |
1388 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1389 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1390 | return p; | |
1da177e4 LT |
1391 | } |
1392 | ||
e2761a11 OH |
1393 | p = d_ancestor(p1, p2); |
1394 | if (p) { | |
1395 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1396 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1397 | return p; | |
1da177e4 LT |
1398 | } |
1399 | ||
f2eace23 IM |
1400 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1401 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1402 | return NULL; |
1403 | } | |
1404 | ||
1405 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1406 | { | |
1b1dcc1b | 1407 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1408 | if (p1 != p2) { |
1b1dcc1b | 1409 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1410 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1411 | } |
1412 | } | |
1413 | ||
1414 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1415 | struct nameidata *nd) | |
1416 | { | |
a95164d9 | 1417 | int error = may_create(dir, dentry); |
1da177e4 LT |
1418 | |
1419 | if (error) | |
1420 | return error; | |
1421 | ||
acfa4380 | 1422 | if (!dir->i_op->create) |
1da177e4 LT |
1423 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1424 | mode &= S_IALLUGO; | |
1425 | mode |= S_IFREG; | |
1426 | error = security_inode_create(dir, dentry, mode); | |
1427 | if (error) | |
1428 | return error; | |
9e3509e2 | 1429 | vfs_dq_init(dir); |
1da177e4 | 1430 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1431 | if (!error) |
f38aa942 | 1432 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1433 | return error; |
1434 | } | |
1435 | ||
3fb64190 | 1436 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1437 | { |
3fb64190 | 1438 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1439 | struct inode *inode = dentry->d_inode; |
1440 | int error; | |
1441 | ||
1442 | if (!inode) | |
1443 | return -ENOENT; | |
1444 | ||
c8fe8f30 CH |
1445 | switch (inode->i_mode & S_IFMT) { |
1446 | case S_IFLNK: | |
1da177e4 | 1447 | return -ELOOP; |
c8fe8f30 CH |
1448 | case S_IFDIR: |
1449 | if (acc_mode & MAY_WRITE) | |
1450 | return -EISDIR; | |
1451 | break; | |
1452 | case S_IFBLK: | |
1453 | case S_IFCHR: | |
3fb64190 | 1454 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1455 | return -EACCES; |
c8fe8f30 CH |
1456 | /*FALLTHRU*/ |
1457 | case S_IFIFO: | |
1458 | case S_IFSOCK: | |
1da177e4 | 1459 | flag &= ~O_TRUNC; |
c8fe8f30 | 1460 | break; |
4a3fd211 | 1461 | } |
b41572e9 | 1462 | |
3fb64190 | 1463 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1464 | if (error) |
1465 | return error; | |
6146f0d5 | 1466 | |
6c1488fd MZ |
1467 | error = ima_path_check(path, acc_mode ? |
1468 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC) : | |
1469 | ACC_MODE(flag) & (MAY_READ | MAY_WRITE), | |
b9fc745d | 1470 | IMA_COUNT_UPDATE); |
6c1488fd | 1471 | |
6146f0d5 MZ |
1472 | if (error) |
1473 | return error; | |
1da177e4 LT |
1474 | /* |
1475 | * An append-only file must be opened in append mode for writing. | |
1476 | */ | |
1477 | if (IS_APPEND(inode)) { | |
acd0c935 | 1478 | error = -EPERM; |
1da177e4 | 1479 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) |
acd0c935 | 1480 | goto err_out; |
1da177e4 | 1481 | if (flag & O_TRUNC) |
acd0c935 | 1482 | goto err_out; |
1da177e4 LT |
1483 | } |
1484 | ||
1485 | /* O_NOATIME can only be set by the owner or superuser */ | |
1486 | if (flag & O_NOATIME) | |
acd0c935 MZ |
1487 | if (!is_owner_or_cap(inode)) { |
1488 | error = -EPERM; | |
1489 | goto err_out; | |
1490 | } | |
1da177e4 LT |
1491 | |
1492 | /* | |
1493 | * Ensure there are no outstanding leases on the file. | |
1494 | */ | |
1495 | error = break_lease(inode, flag); | |
1496 | if (error) | |
acd0c935 | 1497 | goto err_out; |
1da177e4 LT |
1498 | |
1499 | if (flag & O_TRUNC) { | |
1500 | error = get_write_access(inode); | |
1501 | if (error) | |
acd0c935 | 1502 | goto err_out; |
1da177e4 LT |
1503 | |
1504 | /* | |
1505 | * Refuse to truncate files with mandatory locks held on them. | |
1506 | */ | |
1507 | error = locks_verify_locked(inode); | |
be6d3e56 | 1508 | if (!error) |
3fb64190 | 1509 | error = security_path_truncate(path, 0, |
be6d3e56 | 1510 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN); |
1da177e4 | 1511 | if (!error) { |
9e3509e2 | 1512 | vfs_dq_init(inode); |
d139d7ff MS |
1513 | |
1514 | error = do_truncate(dentry, 0, | |
1515 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1516 | NULL); | |
1da177e4 LT |
1517 | } |
1518 | put_write_access(inode); | |
1519 | if (error) | |
acd0c935 | 1520 | goto err_out; |
1da177e4 LT |
1521 | } else |
1522 | if (flag & FMODE_WRITE) | |
9e3509e2 | 1523 | vfs_dq_init(inode); |
1da177e4 LT |
1524 | |
1525 | return 0; | |
acd0c935 MZ |
1526 | err_out: |
1527 | ima_counts_put(path, acc_mode ? | |
1528 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC) : | |
1529 | ACC_MODE(flag) & (MAY_READ | MAY_WRITE)); | |
1530 | return error; | |
1da177e4 LT |
1531 | } |
1532 | ||
d57999e1 DH |
1533 | /* |
1534 | * Be careful about ever adding any more callers of this | |
1535 | * function. Its flags must be in the namei format, not | |
1536 | * what get passed to sys_open(). | |
1537 | */ | |
1538 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1539 | int flag, int mode) |
1540 | { | |
1541 | int error; | |
4ac91378 | 1542 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1543 | |
1544 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1545 | mode &= ~current_umask(); |
be6d3e56 KT |
1546 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1547 | if (error) | |
1548 | goto out_unlock; | |
aab520e2 | 1549 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1550 | out_unlock: |
aab520e2 | 1551 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1552 | dput(nd->path.dentry); |
1553 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1554 | if (error) |
1555 | return error; | |
1556 | /* Don't check for write permission, don't truncate */ | |
3fb64190 | 1557 | return may_open(&nd->path, 0, flag & ~O_TRUNC); |
aab520e2 DH |
1558 | } |
1559 | ||
d57999e1 DH |
1560 | /* |
1561 | * Note that while the flag value (low two bits) for sys_open means: | |
1562 | * 00 - read-only | |
1563 | * 01 - write-only | |
1564 | * 10 - read-write | |
1565 | * 11 - special | |
1566 | * it is changed into | |
1567 | * 00 - no permissions needed | |
1568 | * 01 - read-permission | |
1569 | * 10 - write-permission | |
1570 | * 11 - read-write | |
1571 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1572 | * This is more logical, and also allows the 00 "no perm needed" | |
1573 | * to be used for symlinks (where the permissions are checked | |
1574 | * later). | |
1575 | * | |
1576 | */ | |
1577 | static inline int open_to_namei_flags(int flag) | |
1578 | { | |
1579 | if ((flag+1) & O_ACCMODE) | |
1580 | flag++; | |
1581 | return flag; | |
1582 | } | |
1583 | ||
4a3fd211 DH |
1584 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1585 | { | |
1586 | /* | |
1587 | * We'll never write to the fs underlying | |
1588 | * a device file. | |
1589 | */ | |
1590 | if (special_file(inode->i_mode)) | |
1591 | return 0; | |
1592 | return (flag & O_TRUNC); | |
1593 | } | |
1594 | ||
1da177e4 | 1595 | /* |
4a3fd211 DH |
1596 | * Note that the low bits of the passed in "open_flag" |
1597 | * are not the same as in the local variable "flag". See | |
1598 | * open_to_namei_flags() for more details. | |
1da177e4 | 1599 | */ |
a70e65df | 1600 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1601 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1602 | { |
4a3fd211 | 1603 | struct file *filp; |
a70e65df | 1604 | struct nameidata nd; |
6e8341a1 | 1605 | int error; |
6de88d72 | 1606 | struct path path, save; |
1da177e4 LT |
1607 | struct dentry *dir; |
1608 | int count = 0; | |
4a3fd211 | 1609 | int will_write; |
d57999e1 | 1610 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1611 | |
6b2f3d1f CH |
1612 | /* |
1613 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1614 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1615 | * always set instead of having to deal with possibly weird behaviour | |
1616 | * for malicious applications setting only __O_SYNC. | |
1617 | */ | |
1618 | if (open_flag & __O_SYNC) | |
1619 | open_flag |= O_DSYNC; | |
1620 | ||
6e8341a1 AV |
1621 | if (!acc_mode) |
1622 | acc_mode = MAY_OPEN | ACC_MODE(flag); | |
1da177e4 | 1623 | |
834f2a4a TM |
1624 | /* O_TRUNC implies we need access checks for write permissions */ |
1625 | if (flag & O_TRUNC) | |
1626 | acc_mode |= MAY_WRITE; | |
1627 | ||
1da177e4 LT |
1628 | /* Allow the LSM permission hook to distinguish append |
1629 | access from general write access. */ | |
1630 | if (flag & O_APPEND) | |
1631 | acc_mode |= MAY_APPEND; | |
1632 | ||
1da177e4 LT |
1633 | /* |
1634 | * The simplest case - just a plain lookup. | |
1635 | */ | |
1636 | if (!(flag & O_CREAT)) { | |
2dd6d1f4 AV |
1637 | filp = get_empty_filp(); |
1638 | ||
1639 | if (filp == NULL) | |
1640 | return ERR_PTR(-ENFILE); | |
1641 | nd.intent.open.file = filp; | |
1642 | nd.intent.open.flags = flag; | |
1643 | nd.intent.open.create_mode = 0; | |
1644 | error = do_path_lookup(dfd, pathname, | |
1645 | lookup_flags(flag)|LOOKUP_OPEN, &nd); | |
1646 | if (IS_ERR(nd.intent.open.file)) { | |
1647 | if (error == 0) { | |
1648 | error = PTR_ERR(nd.intent.open.file); | |
1649 | path_put(&nd.path); | |
1650 | } | |
1651 | } else if (error) | |
1652 | release_open_intent(&nd); | |
1da177e4 | 1653 | if (error) |
a70e65df | 1654 | return ERR_PTR(error); |
1da177e4 LT |
1655 | goto ok; |
1656 | } | |
1657 | ||
1658 | /* | |
1659 | * Create - we need to know the parent. | |
1660 | */ | |
9b4a9b14 | 1661 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1662 | if (error) |
a70e65df | 1663 | return ERR_PTR(error); |
9b4a9b14 | 1664 | error = path_walk(pathname, &nd); |
654f562c O |
1665 | if (error) { |
1666 | if (nd.root.mnt) | |
1667 | path_put(&nd.root); | |
9b4a9b14 | 1668 | return ERR_PTR(error); |
654f562c | 1669 | } |
9b4a9b14 AV |
1670 | if (unlikely(!audit_dummy_context())) |
1671 | audit_inode(pathname, nd.path.dentry); | |
1da177e4 LT |
1672 | |
1673 | /* | |
1674 | * We have the parent and last component. First of all, check | |
1675 | * that we are not asked to creat(2) an obvious directory - that | |
1676 | * will not do. | |
1677 | */ | |
1678 | error = -EISDIR; | |
a70e65df | 1679 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1680 | goto exit_parent; |
1da177e4 | 1681 | |
8737f3a1 AV |
1682 | error = -ENFILE; |
1683 | filp = get_empty_filp(); | |
1684 | if (filp == NULL) | |
1685 | goto exit_parent; | |
1686 | nd.intent.open.file = filp; | |
1687 | nd.intent.open.flags = flag; | |
1688 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1689 | dir = nd.path.dentry; |
1690 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1691 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1692 | if (flag & O_EXCL) |
1693 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1694 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1695 | path.dentry = lookup_hash(&nd); |
1696 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1697 | |
1698 | do_last: | |
4e7506e4 AV |
1699 | error = PTR_ERR(path.dentry); |
1700 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1701 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1702 | goto exit; |
1703 | } | |
1704 | ||
a70e65df | 1705 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1706 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1707 | goto exit_mutex_unlock; |
4af4c52f OD |
1708 | } |
1709 | ||
1da177e4 | 1710 | /* Negative dentry, just create the file */ |
4e7506e4 | 1711 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1712 | /* |
1713 | * This write is needed to ensure that a | |
1714 | * ro->rw transition does not occur between | |
1715 | * the time when the file is created and when | |
1716 | * a permanent write count is taken through | |
1717 | * the 'struct file' in nameidata_to_filp(). | |
1718 | */ | |
1719 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1720 | if (error) |
4a3fd211 DH |
1721 | goto exit_mutex_unlock; |
1722 | error = __open_namei_create(&nd, &path, flag, mode); | |
1723 | if (error) { | |
1724 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1725 | goto exit; |
4a3fd211 DH |
1726 | } |
1727 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1728 | if (IS_ERR(filp)) |
1729 | ima_counts_put(&nd.path, | |
1730 | acc_mode & (MAY_READ | MAY_WRITE | | |
1731 | MAY_EXEC)); | |
4a3fd211 | 1732 | mnt_drop_write(nd.path.mnt); |
654f562c O |
1733 | if (nd.root.mnt) |
1734 | path_put(&nd.root); | |
4a3fd211 | 1735 | return filp; |
1da177e4 LT |
1736 | } |
1737 | ||
1738 | /* | |
1739 | * It already exists. | |
1740 | */ | |
1b1dcc1b | 1741 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1742 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1743 | |
1744 | error = -EEXIST; | |
1745 | if (flag & O_EXCL) | |
1746 | goto exit_dput; | |
1747 | ||
e13b210f | 1748 | if (__follow_mount(&path)) { |
1da177e4 | 1749 | error = -ELOOP; |
ba7a4c1a AV |
1750 | if (flag & O_NOFOLLOW) |
1751 | goto exit_dput; | |
1da177e4 | 1752 | } |
3e2efce0 | 1753 | |
1da177e4 | 1754 | error = -ENOENT; |
4e7506e4 | 1755 | if (!path.dentry->d_inode) |
1da177e4 | 1756 | goto exit_dput; |
acfa4380 | 1757 | if (path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1758 | goto do_link; |
1759 | ||
a70e65df | 1760 | path_to_nameidata(&path, &nd); |
1da177e4 | 1761 | error = -EISDIR; |
4e7506e4 | 1762 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1763 | goto exit; |
1764 | ok: | |
4a3fd211 DH |
1765 | /* |
1766 | * Consider: | |
1767 | * 1. may_open() truncates a file | |
1768 | * 2. a rw->ro mount transition occurs | |
1769 | * 3. nameidata_to_filp() fails due to | |
1770 | * the ro mount. | |
1771 | * That would be inconsistent, and should | |
1772 | * be avoided. Taking this mnt write here | |
1773 | * ensures that (2) can not occur. | |
1774 | */ | |
1775 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1776 | if (will_write) { | |
1777 | error = mnt_want_write(nd.path.mnt); | |
1778 | if (error) | |
1779 | goto exit; | |
1780 | } | |
3fb64190 | 1781 | error = may_open(&nd.path, acc_mode, flag); |
4a3fd211 DH |
1782 | if (error) { |
1783 | if (will_write) | |
1784 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1785 | goto exit; |
4a3fd211 DH |
1786 | } |
1787 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1788 | if (IS_ERR(filp)) |
1789 | ima_counts_put(&nd.path, | |
1790 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC)); | |
4a3fd211 DH |
1791 | /* |
1792 | * It is now safe to drop the mnt write | |
1793 | * because the filp has had a write taken | |
1794 | * on its behalf. | |
1795 | */ | |
1796 | if (will_write) | |
1797 | mnt_drop_write(nd.path.mnt); | |
654f562c O |
1798 | if (nd.root.mnt) |
1799 | path_put(&nd.root); | |
4a3fd211 | 1800 | return filp; |
1da177e4 | 1801 | |
4a3fd211 DH |
1802 | exit_mutex_unlock: |
1803 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1804 | exit_dput: |
a70e65df | 1805 | path_put_conditional(&path, &nd); |
1da177e4 | 1806 | exit: |
a70e65df CH |
1807 | if (!IS_ERR(nd.intent.open.file)) |
1808 | release_open_intent(&nd); | |
8737f3a1 | 1809 | exit_parent: |
2a737871 AV |
1810 | if (nd.root.mnt) |
1811 | path_put(&nd.root); | |
a70e65df CH |
1812 | path_put(&nd.path); |
1813 | return ERR_PTR(error); | |
1da177e4 LT |
1814 | |
1815 | do_link: | |
1816 | error = -ELOOP; | |
1817 | if (flag & O_NOFOLLOW) | |
1818 | goto exit_dput; | |
1819 | /* | |
1820 | * This is subtle. Instead of calling do_follow_link() we do the | |
1821 | * thing by hands. The reason is that this way we have zero link_count | |
1822 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1823 | * After that we have the parent and last component, i.e. | |
1824 | * we are in the same situation as after the first path_walk(). | |
1825 | * Well, almost - if the last component is normal we get its copy | |
1826 | * stored in nd->last.name and we will have to putname() it when we | |
1827 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1828 | */ | |
a70e65df CH |
1829 | nd.flags |= LOOKUP_PARENT; |
1830 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1831 | if (error) |
1832 | goto exit_dput; | |
6de88d72 AV |
1833 | save = nd.path; |
1834 | path_get(&save); | |
a70e65df | 1835 | error = __do_follow_link(&path, &nd); |
6de88d72 AV |
1836 | if (error == -ESTALE) { |
1837 | /* nd.path had been dropped */ | |
1838 | nd.path = save; | |
1839 | path_get(&nd.path); | |
1840 | nd.flags |= LOOKUP_REVAL; | |
1841 | error = __do_follow_link(&path, &nd); | |
1842 | } | |
1843 | path_put(&save); | |
258fa999 | 1844 | path_put(&path); |
de459215 KK |
1845 | if (error) { |
1846 | /* Does someone understand code flow here? Or it is only | |
1847 | * me so stupid? Anathema to whoever designed this non-sense | |
1848 | * with "intent.open". | |
1849 | */ | |
a70e65df | 1850 | release_open_intent(&nd); |
654f562c O |
1851 | if (nd.root.mnt) |
1852 | path_put(&nd.root); | |
a70e65df | 1853 | return ERR_PTR(error); |
de459215 | 1854 | } |
a70e65df CH |
1855 | nd.flags &= ~LOOKUP_PARENT; |
1856 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1857 | goto ok; |
1da177e4 | 1858 | error = -EISDIR; |
a70e65df | 1859 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1860 | goto exit; |
a70e65df CH |
1861 | if (nd.last.name[nd.last.len]) { |
1862 | __putname(nd.last.name); | |
1da177e4 LT |
1863 | goto exit; |
1864 | } | |
1865 | error = -ELOOP; | |
1866 | if (count++==32) { | |
a70e65df | 1867 | __putname(nd.last.name); |
1da177e4 LT |
1868 | goto exit; |
1869 | } | |
a70e65df | 1870 | dir = nd.path.dentry; |
1b1dcc1b | 1871 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1872 | path.dentry = lookup_hash(&nd); |
1873 | path.mnt = nd.path.mnt; | |
1874 | __putname(nd.last.name); | |
1da177e4 LT |
1875 | goto do_last; |
1876 | } | |
1877 | ||
a70e65df CH |
1878 | /** |
1879 | * filp_open - open file and return file pointer | |
1880 | * | |
1881 | * @filename: path to open | |
1882 | * @flags: open flags as per the open(2) second argument | |
1883 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1884 | * | |
1885 | * This is the helper to open a file from kernelspace if you really | |
1886 | * have to. But in generally you should not do this, so please move | |
1887 | * along, nothing to see here.. | |
1888 | */ | |
1889 | struct file *filp_open(const char *filename, int flags, int mode) | |
1890 | { | |
6e8341a1 | 1891 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1892 | } |
1893 | EXPORT_SYMBOL(filp_open); | |
1894 | ||
1da177e4 LT |
1895 | /** |
1896 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1897 | * @nd: nameidata info | |
1898 | * @is_dir: directory flag | |
1899 | * | |
1900 | * Simple function to lookup and return a dentry and create it | |
1901 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1902 | * |
4ac91378 | 1903 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1904 | */ |
1905 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1906 | { | |
c663e5d8 | 1907 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1908 | |
4ac91378 | 1909 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1910 | /* |
1911 | * Yucky last component or no last component at all? | |
1912 | * (foo/., foo/.., /////) | |
1913 | */ | |
1da177e4 LT |
1914 | if (nd->last_type != LAST_NORM) |
1915 | goto fail; | |
1916 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1917 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1918 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1919 | |
1920 | /* | |
1921 | * Do the final lookup. | |
1922 | */ | |
49705b77 | 1923 | dentry = lookup_hash(nd); |
1da177e4 LT |
1924 | if (IS_ERR(dentry)) |
1925 | goto fail; | |
c663e5d8 | 1926 | |
e9baf6e5 AV |
1927 | if (dentry->d_inode) |
1928 | goto eexist; | |
c663e5d8 CH |
1929 | /* |
1930 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1931 | * From the vfs_mknod() POV we just have a negative dentry - | |
1932 | * all is fine. Let's be bastards - you had / on the end, you've | |
1933 | * been asking for (non-existent) directory. -ENOENT for you. | |
1934 | */ | |
e9baf6e5 AV |
1935 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1936 | dput(dentry); | |
1937 | dentry = ERR_PTR(-ENOENT); | |
1938 | } | |
1da177e4 | 1939 | return dentry; |
e9baf6e5 | 1940 | eexist: |
1da177e4 | 1941 | dput(dentry); |
e9baf6e5 | 1942 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1943 | fail: |
1944 | return dentry; | |
1945 | } | |
f81a0bff | 1946 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1947 | |
1948 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1949 | { | |
a95164d9 | 1950 | int error = may_create(dir, dentry); |
1da177e4 LT |
1951 | |
1952 | if (error) | |
1953 | return error; | |
1954 | ||
1955 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1956 | return -EPERM; | |
1957 | ||
acfa4380 | 1958 | if (!dir->i_op->mknod) |
1da177e4 LT |
1959 | return -EPERM; |
1960 | ||
08ce5f16 SH |
1961 | error = devcgroup_inode_mknod(mode, dev); |
1962 | if (error) | |
1963 | return error; | |
1964 | ||
1da177e4 LT |
1965 | error = security_inode_mknod(dir, dentry, mode, dev); |
1966 | if (error) | |
1967 | return error; | |
1968 | ||
9e3509e2 | 1969 | vfs_dq_init(dir); |
1da177e4 | 1970 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1971 | if (!error) |
f38aa942 | 1972 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1973 | return error; |
1974 | } | |
1975 | ||
463c3197 DH |
1976 | static int may_mknod(mode_t mode) |
1977 | { | |
1978 | switch (mode & S_IFMT) { | |
1979 | case S_IFREG: | |
1980 | case S_IFCHR: | |
1981 | case S_IFBLK: | |
1982 | case S_IFIFO: | |
1983 | case S_IFSOCK: | |
1984 | case 0: /* zero mode translates to S_IFREG */ | |
1985 | return 0; | |
1986 | case S_IFDIR: | |
1987 | return -EPERM; | |
1988 | default: | |
1989 | return -EINVAL; | |
1990 | } | |
1991 | } | |
1992 | ||
2e4d0924 HC |
1993 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
1994 | unsigned, dev) | |
1da177e4 | 1995 | { |
2ad94ae6 AV |
1996 | int error; |
1997 | char *tmp; | |
1998 | struct dentry *dentry; | |
1da177e4 LT |
1999 | struct nameidata nd; |
2000 | ||
2001 | if (S_ISDIR(mode)) | |
2002 | return -EPERM; | |
1da177e4 | 2003 | |
2ad94ae6 | 2004 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2005 | if (error) |
2ad94ae6 AV |
2006 | return error; |
2007 | ||
1da177e4 | 2008 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2009 | if (IS_ERR(dentry)) { |
2010 | error = PTR_ERR(dentry); | |
2011 | goto out_unlock; | |
2012 | } | |
4ac91378 | 2013 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2014 | mode &= ~current_umask(); |
463c3197 DH |
2015 | error = may_mknod(mode); |
2016 | if (error) | |
2017 | goto out_dput; | |
2018 | error = mnt_want_write(nd.path.mnt); | |
2019 | if (error) | |
2020 | goto out_dput; | |
be6d3e56 KT |
2021 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2022 | if (error) | |
2023 | goto out_drop_write; | |
463c3197 | 2024 | switch (mode & S_IFMT) { |
1da177e4 | 2025 | case 0: case S_IFREG: |
4ac91378 | 2026 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2027 | break; |
2028 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2029 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2030 | new_decode_dev(dev)); |
2031 | break; | |
2032 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2033 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2034 | break; |
1da177e4 | 2035 | } |
be6d3e56 | 2036 | out_drop_write: |
463c3197 DH |
2037 | mnt_drop_write(nd.path.mnt); |
2038 | out_dput: | |
2039 | dput(dentry); | |
2040 | out_unlock: | |
4ac91378 | 2041 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2042 | path_put(&nd.path); |
1da177e4 LT |
2043 | putname(tmp); |
2044 | ||
2045 | return error; | |
2046 | } | |
2047 | ||
3480b257 | 2048 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2049 | { |
2050 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2051 | } | |
2052 | ||
1da177e4 LT |
2053 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2054 | { | |
a95164d9 | 2055 | int error = may_create(dir, dentry); |
1da177e4 LT |
2056 | |
2057 | if (error) | |
2058 | return error; | |
2059 | ||
acfa4380 | 2060 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2061 | return -EPERM; |
2062 | ||
2063 | mode &= (S_IRWXUGO|S_ISVTX); | |
2064 | error = security_inode_mkdir(dir, dentry, mode); | |
2065 | if (error) | |
2066 | return error; | |
2067 | ||
9e3509e2 | 2068 | vfs_dq_init(dir); |
1da177e4 | 2069 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2070 | if (!error) |
f38aa942 | 2071 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2072 | return error; |
2073 | } | |
2074 | ||
2e4d0924 | 2075 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2076 | { |
2077 | int error = 0; | |
2078 | char * tmp; | |
6902d925 DH |
2079 | struct dentry *dentry; |
2080 | struct nameidata nd; | |
1da177e4 | 2081 | |
2ad94ae6 AV |
2082 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2083 | if (error) | |
6902d925 | 2084 | goto out_err; |
1da177e4 | 2085 | |
6902d925 DH |
2086 | dentry = lookup_create(&nd, 1); |
2087 | error = PTR_ERR(dentry); | |
2088 | if (IS_ERR(dentry)) | |
2089 | goto out_unlock; | |
1da177e4 | 2090 | |
4ac91378 | 2091 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2092 | mode &= ~current_umask(); |
463c3197 DH |
2093 | error = mnt_want_write(nd.path.mnt); |
2094 | if (error) | |
2095 | goto out_dput; | |
be6d3e56 KT |
2096 | error = security_path_mkdir(&nd.path, dentry, mode); |
2097 | if (error) | |
2098 | goto out_drop_write; | |
4ac91378 | 2099 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2100 | out_drop_write: |
463c3197 DH |
2101 | mnt_drop_write(nd.path.mnt); |
2102 | out_dput: | |
6902d925 DH |
2103 | dput(dentry); |
2104 | out_unlock: | |
4ac91378 | 2105 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2106 | path_put(&nd.path); |
6902d925 DH |
2107 | putname(tmp); |
2108 | out_err: | |
1da177e4 LT |
2109 | return error; |
2110 | } | |
2111 | ||
3cdad428 | 2112 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2113 | { |
2114 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2115 | } | |
2116 | ||
1da177e4 LT |
2117 | /* |
2118 | * We try to drop the dentry early: we should have | |
2119 | * a usage count of 2 if we're the only user of this | |
2120 | * dentry, and if that is true (possibly after pruning | |
2121 | * the dcache), then we drop the dentry now. | |
2122 | * | |
2123 | * A low-level filesystem can, if it choses, legally | |
2124 | * do a | |
2125 | * | |
2126 | * if (!d_unhashed(dentry)) | |
2127 | * return -EBUSY; | |
2128 | * | |
2129 | * if it cannot handle the case of removing a directory | |
2130 | * that is still in use by something else.. | |
2131 | */ | |
2132 | void dentry_unhash(struct dentry *dentry) | |
2133 | { | |
2134 | dget(dentry); | |
dc168427 | 2135 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2136 | spin_lock(&dcache_lock); |
2137 | spin_lock(&dentry->d_lock); | |
2138 | if (atomic_read(&dentry->d_count) == 2) | |
2139 | __d_drop(dentry); | |
2140 | spin_unlock(&dentry->d_lock); | |
2141 | spin_unlock(&dcache_lock); | |
2142 | } | |
2143 | ||
2144 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2145 | { | |
2146 | int error = may_delete(dir, dentry, 1); | |
2147 | ||
2148 | if (error) | |
2149 | return error; | |
2150 | ||
acfa4380 | 2151 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2152 | return -EPERM; |
2153 | ||
9e3509e2 | 2154 | vfs_dq_init(dir); |
1da177e4 | 2155 | |
1b1dcc1b | 2156 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2157 | dentry_unhash(dentry); |
2158 | if (d_mountpoint(dentry)) | |
2159 | error = -EBUSY; | |
2160 | else { | |
2161 | error = security_inode_rmdir(dir, dentry); | |
2162 | if (!error) { | |
2163 | error = dir->i_op->rmdir(dir, dentry); | |
2164 | if (!error) | |
2165 | dentry->d_inode->i_flags |= S_DEAD; | |
2166 | } | |
2167 | } | |
1b1dcc1b | 2168 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2169 | if (!error) { |
1da177e4 LT |
2170 | d_delete(dentry); |
2171 | } | |
2172 | dput(dentry); | |
2173 | ||
2174 | return error; | |
2175 | } | |
2176 | ||
5590ff0d | 2177 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2178 | { |
2179 | int error = 0; | |
2180 | char * name; | |
2181 | struct dentry *dentry; | |
2182 | struct nameidata nd; | |
2183 | ||
2ad94ae6 | 2184 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2185 | if (error) |
2ad94ae6 | 2186 | return error; |
1da177e4 LT |
2187 | |
2188 | switch(nd.last_type) { | |
0612d9fb OH |
2189 | case LAST_DOTDOT: |
2190 | error = -ENOTEMPTY; | |
2191 | goto exit1; | |
2192 | case LAST_DOT: | |
2193 | error = -EINVAL; | |
2194 | goto exit1; | |
2195 | case LAST_ROOT: | |
2196 | error = -EBUSY; | |
2197 | goto exit1; | |
1da177e4 | 2198 | } |
0612d9fb OH |
2199 | |
2200 | nd.flags &= ~LOOKUP_PARENT; | |
2201 | ||
4ac91378 | 2202 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2203 | dentry = lookup_hash(&nd); |
1da177e4 | 2204 | error = PTR_ERR(dentry); |
6902d925 DH |
2205 | if (IS_ERR(dentry)) |
2206 | goto exit2; | |
0622753b DH |
2207 | error = mnt_want_write(nd.path.mnt); |
2208 | if (error) | |
2209 | goto exit3; | |
be6d3e56 KT |
2210 | error = security_path_rmdir(&nd.path, dentry); |
2211 | if (error) | |
2212 | goto exit4; | |
4ac91378 | 2213 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2214 | exit4: |
0622753b DH |
2215 | mnt_drop_write(nd.path.mnt); |
2216 | exit3: | |
6902d925 DH |
2217 | dput(dentry); |
2218 | exit2: | |
4ac91378 | 2219 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2220 | exit1: |
1d957f9b | 2221 | path_put(&nd.path); |
1da177e4 LT |
2222 | putname(name); |
2223 | return error; | |
2224 | } | |
2225 | ||
3cdad428 | 2226 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2227 | { |
2228 | return do_rmdir(AT_FDCWD, pathname); | |
2229 | } | |
2230 | ||
1da177e4 LT |
2231 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2232 | { | |
2233 | int error = may_delete(dir, dentry, 0); | |
2234 | ||
2235 | if (error) | |
2236 | return error; | |
2237 | ||
acfa4380 | 2238 | if (!dir->i_op->unlink) |
1da177e4 LT |
2239 | return -EPERM; |
2240 | ||
9e3509e2 | 2241 | vfs_dq_init(dir); |
1da177e4 | 2242 | |
1b1dcc1b | 2243 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2244 | if (d_mountpoint(dentry)) |
2245 | error = -EBUSY; | |
2246 | else { | |
2247 | error = security_inode_unlink(dir, dentry); | |
2248 | if (!error) | |
2249 | error = dir->i_op->unlink(dir, dentry); | |
2250 | } | |
1b1dcc1b | 2251 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2252 | |
2253 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2254 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2255 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2256 | d_delete(dentry); |
1da177e4 | 2257 | } |
0eeca283 | 2258 | |
1da177e4 LT |
2259 | return error; |
2260 | } | |
2261 | ||
2262 | /* | |
2263 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2264 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2265 | * writeout happening, and we don't want to prevent access to the directory |
2266 | * while waiting on the I/O. | |
2267 | */ | |
5590ff0d | 2268 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2269 | { |
2ad94ae6 AV |
2270 | int error; |
2271 | char *name; | |
1da177e4 LT |
2272 | struct dentry *dentry; |
2273 | struct nameidata nd; | |
2274 | struct inode *inode = NULL; | |
2275 | ||
2ad94ae6 | 2276 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2277 | if (error) |
2ad94ae6 AV |
2278 | return error; |
2279 | ||
1da177e4 LT |
2280 | error = -EISDIR; |
2281 | if (nd.last_type != LAST_NORM) | |
2282 | goto exit1; | |
0612d9fb OH |
2283 | |
2284 | nd.flags &= ~LOOKUP_PARENT; | |
2285 | ||
4ac91378 | 2286 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2287 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2288 | error = PTR_ERR(dentry); |
2289 | if (!IS_ERR(dentry)) { | |
2290 | /* Why not before? Because we want correct error value */ | |
2291 | if (nd.last.name[nd.last.len]) | |
2292 | goto slashes; | |
2293 | inode = dentry->d_inode; | |
2294 | if (inode) | |
2295 | atomic_inc(&inode->i_count); | |
0622753b DH |
2296 | error = mnt_want_write(nd.path.mnt); |
2297 | if (error) | |
2298 | goto exit2; | |
be6d3e56 KT |
2299 | error = security_path_unlink(&nd.path, dentry); |
2300 | if (error) | |
2301 | goto exit3; | |
4ac91378 | 2302 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2303 | exit3: |
0622753b | 2304 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2305 | exit2: |
2306 | dput(dentry); | |
2307 | } | |
4ac91378 | 2308 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2309 | if (inode) |
2310 | iput(inode); /* truncate the inode here */ | |
2311 | exit1: | |
1d957f9b | 2312 | path_put(&nd.path); |
1da177e4 LT |
2313 | putname(name); |
2314 | return error; | |
2315 | ||
2316 | slashes: | |
2317 | error = !dentry->d_inode ? -ENOENT : | |
2318 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2319 | goto exit2; | |
2320 | } | |
2321 | ||
2e4d0924 | 2322 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2323 | { |
2324 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2325 | return -EINVAL; | |
2326 | ||
2327 | if (flag & AT_REMOVEDIR) | |
2328 | return do_rmdir(dfd, pathname); | |
2329 | ||
2330 | return do_unlinkat(dfd, pathname); | |
2331 | } | |
2332 | ||
3480b257 | 2333 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2334 | { |
2335 | return do_unlinkat(AT_FDCWD, pathname); | |
2336 | } | |
2337 | ||
db2e747b | 2338 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2339 | { |
a95164d9 | 2340 | int error = may_create(dir, dentry); |
1da177e4 LT |
2341 | |
2342 | if (error) | |
2343 | return error; | |
2344 | ||
acfa4380 | 2345 | if (!dir->i_op->symlink) |
1da177e4 LT |
2346 | return -EPERM; |
2347 | ||
2348 | error = security_inode_symlink(dir, dentry, oldname); | |
2349 | if (error) | |
2350 | return error; | |
2351 | ||
9e3509e2 | 2352 | vfs_dq_init(dir); |
1da177e4 | 2353 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2354 | if (!error) |
f38aa942 | 2355 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2356 | return error; |
2357 | } | |
2358 | ||
2e4d0924 HC |
2359 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2360 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2361 | { |
2ad94ae6 AV |
2362 | int error; |
2363 | char *from; | |
2364 | char *to; | |
6902d925 DH |
2365 | struct dentry *dentry; |
2366 | struct nameidata nd; | |
1da177e4 LT |
2367 | |
2368 | from = getname(oldname); | |
2ad94ae6 | 2369 | if (IS_ERR(from)) |
1da177e4 | 2370 | return PTR_ERR(from); |
1da177e4 | 2371 | |
2ad94ae6 | 2372 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2373 | if (error) |
2ad94ae6 AV |
2374 | goto out_putname; |
2375 | ||
6902d925 DH |
2376 | dentry = lookup_create(&nd, 0); |
2377 | error = PTR_ERR(dentry); | |
2378 | if (IS_ERR(dentry)) | |
2379 | goto out_unlock; | |
2380 | ||
75c3f29d DH |
2381 | error = mnt_want_write(nd.path.mnt); |
2382 | if (error) | |
2383 | goto out_dput; | |
be6d3e56 KT |
2384 | error = security_path_symlink(&nd.path, dentry, from); |
2385 | if (error) | |
2386 | goto out_drop_write; | |
db2e747b | 2387 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2388 | out_drop_write: |
75c3f29d DH |
2389 | mnt_drop_write(nd.path.mnt); |
2390 | out_dput: | |
6902d925 DH |
2391 | dput(dentry); |
2392 | out_unlock: | |
4ac91378 | 2393 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2394 | path_put(&nd.path); |
6902d925 DH |
2395 | putname(to); |
2396 | out_putname: | |
1da177e4 LT |
2397 | putname(from); |
2398 | return error; | |
2399 | } | |
2400 | ||
3480b257 | 2401 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2402 | { |
2403 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2404 | } | |
2405 | ||
1da177e4 LT |
2406 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2407 | { | |
2408 | struct inode *inode = old_dentry->d_inode; | |
2409 | int error; | |
2410 | ||
2411 | if (!inode) | |
2412 | return -ENOENT; | |
2413 | ||
a95164d9 | 2414 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2415 | if (error) |
2416 | return error; | |
2417 | ||
2418 | if (dir->i_sb != inode->i_sb) | |
2419 | return -EXDEV; | |
2420 | ||
2421 | /* | |
2422 | * A link to an append-only or immutable file cannot be created. | |
2423 | */ | |
2424 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2425 | return -EPERM; | |
acfa4380 | 2426 | if (!dir->i_op->link) |
1da177e4 | 2427 | return -EPERM; |
7e79eedb | 2428 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2429 | return -EPERM; |
2430 | ||
2431 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2432 | if (error) | |
2433 | return error; | |
2434 | ||
7e79eedb | 2435 | mutex_lock(&inode->i_mutex); |
9e3509e2 | 2436 | vfs_dq_init(dir); |
1da177e4 | 2437 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2438 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2439 | if (!error) |
7e79eedb | 2440 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2441 | return error; |
2442 | } | |
2443 | ||
2444 | /* | |
2445 | * Hardlinks are often used in delicate situations. We avoid | |
2446 | * security-related surprises by not following symlinks on the | |
2447 | * newname. --KAB | |
2448 | * | |
2449 | * We don't follow them on the oldname either to be compatible | |
2450 | * with linux 2.0, and to avoid hard-linking to directories | |
2451 | * and other special files. --ADM | |
2452 | */ | |
2e4d0924 HC |
2453 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2454 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2455 | { |
2456 | struct dentry *new_dentry; | |
2d8f3038 AV |
2457 | struct nameidata nd; |
2458 | struct path old_path; | |
1da177e4 | 2459 | int error; |
2ad94ae6 | 2460 | char *to; |
1da177e4 | 2461 | |
45c9b11a | 2462 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2463 | return -EINVAL; |
2464 | ||
2d8f3038 AV |
2465 | error = user_path_at(olddfd, oldname, |
2466 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2467 | &old_path); | |
1da177e4 | 2468 | if (error) |
2ad94ae6 AV |
2469 | return error; |
2470 | ||
2471 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2472 | if (error) |
2473 | goto out; | |
2474 | error = -EXDEV; | |
2d8f3038 | 2475 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2476 | goto out_release; |
2477 | new_dentry = lookup_create(&nd, 0); | |
2478 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2479 | if (IS_ERR(new_dentry)) |
2480 | goto out_unlock; | |
75c3f29d DH |
2481 | error = mnt_want_write(nd.path.mnt); |
2482 | if (error) | |
2483 | goto out_dput; | |
be6d3e56 KT |
2484 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2485 | if (error) | |
2486 | goto out_drop_write; | |
2d8f3038 | 2487 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2488 | out_drop_write: |
75c3f29d DH |
2489 | mnt_drop_write(nd.path.mnt); |
2490 | out_dput: | |
6902d925 DH |
2491 | dput(new_dentry); |
2492 | out_unlock: | |
4ac91378 | 2493 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2494 | out_release: |
1d957f9b | 2495 | path_put(&nd.path); |
2ad94ae6 | 2496 | putname(to); |
1da177e4 | 2497 | out: |
2d8f3038 | 2498 | path_put(&old_path); |
1da177e4 LT |
2499 | |
2500 | return error; | |
2501 | } | |
2502 | ||
3480b257 | 2503 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2504 | { |
c04030e1 | 2505 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2506 | } |
2507 | ||
1da177e4 LT |
2508 | /* |
2509 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2510 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2511 | * Problems: | |
2512 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2513 | * b) race potential - two innocent renames can create a loop together. | |
2514 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2515 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2516 | * story. |
2517 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2518 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2519 | * whether the target exists). Solution: try to be smart with locking |
2520 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2521 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2522 | * move will be locked. Thus we can rank directories by the tree |
2523 | * (ancestors first) and rank all non-directories after them. | |
2524 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2525 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2526 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2527 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2528 | * we'd better make sure that there's no link(2) for them. | |
2529 | * d) some filesystems don't support opened-but-unlinked directories, | |
2530 | * either because of layout or because they are not ready to deal with | |
2531 | * all cases correctly. The latter will be fixed (taking this sort of | |
2532 | * stuff into VFS), but the former is not going away. Solution: the same | |
2533 | * trick as in rmdir(). | |
2534 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2535 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2536 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2537 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2538 | * locking]. |
2539 | */ | |
75c96f85 AB |
2540 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2541 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2542 | { |
2543 | int error = 0; | |
2544 | struct inode *target; | |
2545 | ||
2546 | /* | |
2547 | * If we are going to change the parent - check write permissions, | |
2548 | * we'll need to flip '..'. | |
2549 | */ | |
2550 | if (new_dir != old_dir) { | |
f419a2e3 | 2551 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2552 | if (error) |
2553 | return error; | |
2554 | } | |
2555 | ||
2556 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2557 | if (error) | |
2558 | return error; | |
2559 | ||
2560 | target = new_dentry->d_inode; | |
2561 | if (target) { | |
1b1dcc1b | 2562 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2563 | dentry_unhash(new_dentry); |
2564 | } | |
2565 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2566 | error = -EBUSY; | |
2567 | else | |
2568 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2569 | if (target) { | |
2570 | if (!error) | |
2571 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2572 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2573 | if (d_unhashed(new_dentry)) |
2574 | d_rehash(new_dentry); | |
2575 | dput(new_dentry); | |
2576 | } | |
e31e14ec | 2577 | if (!error) |
349457cc MF |
2578 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2579 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2580 | return error; |
2581 | } | |
2582 | ||
75c96f85 AB |
2583 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2584 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2585 | { |
2586 | struct inode *target; | |
2587 | int error; | |
2588 | ||
2589 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2590 | if (error) | |
2591 | return error; | |
2592 | ||
2593 | dget(new_dentry); | |
2594 | target = new_dentry->d_inode; | |
2595 | if (target) | |
1b1dcc1b | 2596 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2597 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2598 | error = -EBUSY; | |
2599 | else | |
2600 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2601 | if (!error) { | |
349457cc | 2602 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2603 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2604 | } |
2605 | if (target) | |
1b1dcc1b | 2606 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2607 | dput(new_dentry); |
2608 | return error; | |
2609 | } | |
2610 | ||
2611 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2612 | struct inode *new_dir, struct dentry *new_dentry) | |
2613 | { | |
2614 | int error; | |
2615 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2616 | const char *old_name; |
1da177e4 LT |
2617 | |
2618 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2619 | return 0; | |
2620 | ||
2621 | error = may_delete(old_dir, old_dentry, is_dir); | |
2622 | if (error) | |
2623 | return error; | |
2624 | ||
2625 | if (!new_dentry->d_inode) | |
a95164d9 | 2626 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2627 | else |
2628 | error = may_delete(new_dir, new_dentry, is_dir); | |
2629 | if (error) | |
2630 | return error; | |
2631 | ||
acfa4380 | 2632 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2633 | return -EPERM; |
2634 | ||
9e3509e2 JK |
2635 | vfs_dq_init(old_dir); |
2636 | vfs_dq_init(new_dir); | |
1da177e4 | 2637 | |
0eeca283 RL |
2638 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2639 | ||
1da177e4 LT |
2640 | if (is_dir) |
2641 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2642 | else | |
2643 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2644 | if (!error) { | |
0eeca283 | 2645 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2646 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2647 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2648 | } |
0eeca283 RL |
2649 | fsnotify_oldname_free(old_name); |
2650 | ||
1da177e4 LT |
2651 | return error; |
2652 | } | |
2653 | ||
2e4d0924 HC |
2654 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2655 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2656 | { |
2ad94ae6 AV |
2657 | struct dentry *old_dir, *new_dir; |
2658 | struct dentry *old_dentry, *new_dentry; | |
2659 | struct dentry *trap; | |
1da177e4 | 2660 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2661 | char *from; |
2662 | char *to; | |
2663 | int error; | |
1da177e4 | 2664 | |
2ad94ae6 | 2665 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2666 | if (error) |
2667 | goto exit; | |
2668 | ||
2ad94ae6 | 2669 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2670 | if (error) |
2671 | goto exit1; | |
2672 | ||
2673 | error = -EXDEV; | |
4ac91378 | 2674 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2675 | goto exit2; |
2676 | ||
4ac91378 | 2677 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2678 | error = -EBUSY; |
2679 | if (oldnd.last_type != LAST_NORM) | |
2680 | goto exit2; | |
2681 | ||
4ac91378 | 2682 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2683 | if (newnd.last_type != LAST_NORM) |
2684 | goto exit2; | |
2685 | ||
0612d9fb OH |
2686 | oldnd.flags &= ~LOOKUP_PARENT; |
2687 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2688 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2689 | |
1da177e4 LT |
2690 | trap = lock_rename(new_dir, old_dir); |
2691 | ||
49705b77 | 2692 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2693 | error = PTR_ERR(old_dentry); |
2694 | if (IS_ERR(old_dentry)) | |
2695 | goto exit3; | |
2696 | /* source must exist */ | |
2697 | error = -ENOENT; | |
2698 | if (!old_dentry->d_inode) | |
2699 | goto exit4; | |
2700 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2701 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2702 | error = -ENOTDIR; | |
2703 | if (oldnd.last.name[oldnd.last.len]) | |
2704 | goto exit4; | |
2705 | if (newnd.last.name[newnd.last.len]) | |
2706 | goto exit4; | |
2707 | } | |
2708 | /* source should not be ancestor of target */ | |
2709 | error = -EINVAL; | |
2710 | if (old_dentry == trap) | |
2711 | goto exit4; | |
49705b77 | 2712 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2713 | error = PTR_ERR(new_dentry); |
2714 | if (IS_ERR(new_dentry)) | |
2715 | goto exit4; | |
2716 | /* target should not be an ancestor of source */ | |
2717 | error = -ENOTEMPTY; | |
2718 | if (new_dentry == trap) | |
2719 | goto exit5; | |
2720 | ||
9079b1eb DH |
2721 | error = mnt_want_write(oldnd.path.mnt); |
2722 | if (error) | |
2723 | goto exit5; | |
be6d3e56 KT |
2724 | error = security_path_rename(&oldnd.path, old_dentry, |
2725 | &newnd.path, new_dentry); | |
2726 | if (error) | |
2727 | goto exit6; | |
1da177e4 LT |
2728 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2729 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2730 | exit6: |
9079b1eb | 2731 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2732 | exit5: |
2733 | dput(new_dentry); | |
2734 | exit4: | |
2735 | dput(old_dentry); | |
2736 | exit3: | |
2737 | unlock_rename(new_dir, old_dir); | |
2738 | exit2: | |
1d957f9b | 2739 | path_put(&newnd.path); |
2ad94ae6 | 2740 | putname(to); |
1da177e4 | 2741 | exit1: |
1d957f9b | 2742 | path_put(&oldnd.path); |
1da177e4 | 2743 | putname(from); |
2ad94ae6 | 2744 | exit: |
1da177e4 LT |
2745 | return error; |
2746 | } | |
2747 | ||
a26eab24 | 2748 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2749 | { |
2750 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2751 | } | |
2752 | ||
1da177e4 LT |
2753 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2754 | { | |
2755 | int len; | |
2756 | ||
2757 | len = PTR_ERR(link); | |
2758 | if (IS_ERR(link)) | |
2759 | goto out; | |
2760 | ||
2761 | len = strlen(link); | |
2762 | if (len > (unsigned) buflen) | |
2763 | len = buflen; | |
2764 | if (copy_to_user(buffer, link, len)) | |
2765 | len = -EFAULT; | |
2766 | out: | |
2767 | return len; | |
2768 | } | |
2769 | ||
2770 | /* | |
2771 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2772 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2773 | * using) it for any given inode is up to filesystem. | |
2774 | */ | |
2775 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2776 | { | |
2777 | struct nameidata nd; | |
cc314eef | 2778 | void *cookie; |
694a1764 | 2779 | int res; |
cc314eef | 2780 | |
1da177e4 | 2781 | nd.depth = 0; |
cc314eef | 2782 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2783 | if (IS_ERR(cookie)) |
2784 | return PTR_ERR(cookie); | |
2785 | ||
2786 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2787 | if (dentry->d_inode->i_op->put_link) | |
2788 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2789 | return res; | |
1da177e4 LT |
2790 | } |
2791 | ||
2792 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2793 | { | |
2794 | return __vfs_follow_link(nd, link); | |
2795 | } | |
2796 | ||
2797 | /* get the link contents into pagecache */ | |
2798 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2799 | { | |
ebd09abb DG |
2800 | char *kaddr; |
2801 | struct page *page; | |
1da177e4 | 2802 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2803 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2804 | if (IS_ERR(page)) |
6fe6900e | 2805 | return (char*)page; |
1da177e4 | 2806 | *ppage = page; |
ebd09abb DG |
2807 | kaddr = kmap(page); |
2808 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2809 | return kaddr; | |
1da177e4 LT |
2810 | } |
2811 | ||
2812 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2813 | { | |
2814 | struct page *page = NULL; | |
2815 | char *s = page_getlink(dentry, &page); | |
2816 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2817 | if (page) { | |
2818 | kunmap(page); | |
2819 | page_cache_release(page); | |
2820 | } | |
2821 | return res; | |
2822 | } | |
2823 | ||
cc314eef | 2824 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2825 | { |
cc314eef | 2826 | struct page *page = NULL; |
1da177e4 | 2827 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2828 | return page; |
1da177e4 LT |
2829 | } |
2830 | ||
cc314eef | 2831 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2832 | { |
cc314eef LT |
2833 | struct page *page = cookie; |
2834 | ||
2835 | if (page) { | |
1da177e4 LT |
2836 | kunmap(page); |
2837 | page_cache_release(page); | |
1da177e4 LT |
2838 | } |
2839 | } | |
2840 | ||
54566b2c NP |
2841 | /* |
2842 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2843 | */ | |
2844 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2845 | { |
2846 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2847 | struct page *page; |
afddba49 | 2848 | void *fsdata; |
beb497ab | 2849 | int err; |
1da177e4 | 2850 | char *kaddr; |
54566b2c NP |
2851 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2852 | if (nofs) | |
2853 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2854 | |
7e53cac4 | 2855 | retry: |
afddba49 | 2856 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2857 | flags, &page, &fsdata); |
1da177e4 | 2858 | if (err) |
afddba49 NP |
2859 | goto fail; |
2860 | ||
1da177e4 LT |
2861 | kaddr = kmap_atomic(page, KM_USER0); |
2862 | memcpy(kaddr, symname, len-1); | |
2863 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2864 | |
2865 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2866 | page, fsdata); | |
1da177e4 LT |
2867 | if (err < 0) |
2868 | goto fail; | |
afddba49 NP |
2869 | if (err < len-1) |
2870 | goto retry; | |
2871 | ||
1da177e4 LT |
2872 | mark_inode_dirty(inode); |
2873 | return 0; | |
1da177e4 LT |
2874 | fail: |
2875 | return err; | |
2876 | } | |
2877 | ||
0adb25d2 KK |
2878 | int page_symlink(struct inode *inode, const char *symname, int len) |
2879 | { | |
2880 | return __page_symlink(inode, symname, len, | |
54566b2c | 2881 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2882 | } |
2883 | ||
92e1d5be | 2884 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2885 | .readlink = generic_readlink, |
2886 | .follow_link = page_follow_link_light, | |
2887 | .put_link = page_put_link, | |
2888 | }; | |
2889 | ||
2d8f3038 | 2890 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2891 | EXPORT_SYMBOL(follow_down); |
2892 | EXPORT_SYMBOL(follow_up); | |
2893 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2894 | EXPORT_SYMBOL(getname); | |
2895 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2896 | EXPORT_SYMBOL(lookup_one_len); |
2897 | EXPORT_SYMBOL(page_follow_link_light); | |
2898 | EXPORT_SYMBOL(page_put_link); | |
2899 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2900 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2901 | EXPORT_SYMBOL(page_symlink); |
2902 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2903 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2904 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2905 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2906 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2907 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2908 | EXPORT_SYMBOL(unlock_rename); |
2909 | EXPORT_SYMBOL(vfs_create); | |
2910 | EXPORT_SYMBOL(vfs_follow_link); | |
2911 | EXPORT_SYMBOL(vfs_link); | |
2912 | EXPORT_SYMBOL(vfs_mkdir); | |
2913 | EXPORT_SYMBOL(vfs_mknod); | |
2914 | EXPORT_SYMBOL(generic_permission); | |
2915 | EXPORT_SYMBOL(vfs_readlink); | |
2916 | EXPORT_SYMBOL(vfs_rename); | |
2917 | EXPORT_SYMBOL(vfs_rmdir); | |
2918 | EXPORT_SYMBOL(vfs_symlink); | |
2919 | EXPORT_SYMBOL(vfs_unlink); | |
2920 | EXPORT_SYMBOL(dentry_unhash); | |
2921 | EXPORT_SYMBOL(generic_readlink); |