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 | ||
a02f76c3 | 111 | static int __link_path_walk(const char *name, struct nameidata *nd); |
c4a7808f | 112 | |
1da177e4 LT |
113 | /* In order to reduce some races, while at the same time doing additional |
114 | * checking and hopefully speeding things up, we copy filenames to the | |
115 | * kernel data space before using them.. | |
116 | * | |
117 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
118 | * PATH_MAX includes the nul terminator --RR. | |
119 | */ | |
858119e1 | 120 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
121 | { |
122 | int retval; | |
123 | unsigned long len = PATH_MAX; | |
124 | ||
125 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
126 | if ((unsigned long) filename >= TASK_SIZE) | |
127 | return -EFAULT; | |
128 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
129 | len = TASK_SIZE - (unsigned long) filename; | |
130 | } | |
131 | ||
132 | retval = strncpy_from_user(page, filename, len); | |
133 | if (retval > 0) { | |
134 | if (retval < len) | |
135 | return 0; | |
136 | return -ENAMETOOLONG; | |
137 | } else if (!retval) | |
138 | retval = -ENOENT; | |
139 | return retval; | |
140 | } | |
141 | ||
142 | char * getname(const char __user * filename) | |
143 | { | |
144 | char *tmp, *result; | |
145 | ||
146 | result = ERR_PTR(-ENOMEM); | |
147 | tmp = __getname(); | |
148 | if (tmp) { | |
149 | int retval = do_getname(filename, tmp); | |
150 | ||
151 | result = tmp; | |
152 | if (retval < 0) { | |
153 | __putname(tmp); | |
154 | result = ERR_PTR(retval); | |
155 | } | |
156 | } | |
157 | audit_getname(result); | |
158 | return result; | |
159 | } | |
160 | ||
161 | #ifdef CONFIG_AUDITSYSCALL | |
162 | void putname(const char *name) | |
163 | { | |
5ac3a9c2 | 164 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
165 | audit_putname(name); |
166 | else | |
167 | __putname(name); | |
168 | } | |
169 | EXPORT_SYMBOL(putname); | |
170 | #endif | |
171 | ||
172 | ||
173 | /** | |
174 | * generic_permission - check for access rights on a Posix-like filesystem | |
175 | * @inode: inode to check access rights for | |
176 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
177 | * @check_acl: optional callback to check for Posix ACLs | |
178 | * | |
179 | * Used to check for read/write/execute permissions on a file. | |
180 | * We use "fsuid" for this, letting us set arbitrary permissions | |
181 | * for filesystem access without changing the "normal" uids which | |
182 | * are used for other things.. | |
183 | */ | |
184 | int generic_permission(struct inode *inode, int mask, | |
185 | int (*check_acl)(struct inode *inode, int mask)) | |
186 | { | |
187 | umode_t mode = inode->i_mode; | |
188 | ||
e6305c43 AV |
189 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
190 | ||
da9592ed | 191 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
192 | mode >>= 6; |
193 | else { | |
194 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
195 | int error = check_acl(inode, mask); | |
196 | if (error == -EACCES) | |
197 | goto check_capabilities; | |
198 | else if (error != -EAGAIN) | |
199 | return error; | |
200 | } | |
201 | ||
202 | if (in_group_p(inode->i_gid)) | |
203 | mode >>= 3; | |
204 | } | |
205 | ||
206 | /* | |
207 | * If the DACs are ok we don't need any capability check. | |
208 | */ | |
e6305c43 | 209 | if ((mask & ~mode) == 0) |
1da177e4 LT |
210 | return 0; |
211 | ||
212 | check_capabilities: | |
213 | /* | |
214 | * Read/write DACs are always overridable. | |
215 | * Executable DACs are overridable if at least one exec bit is set. | |
216 | */ | |
f696a365 | 217 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
218 | if (capable(CAP_DAC_OVERRIDE)) |
219 | return 0; | |
220 | ||
221 | /* | |
222 | * Searching includes executable on directories, else just read. | |
223 | */ | |
224 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
225 | if (capable(CAP_DAC_READ_SEARCH)) | |
226 | return 0; | |
227 | ||
228 | return -EACCES; | |
229 | } | |
230 | ||
cb23beb5 CH |
231 | /** |
232 | * inode_permission - check for access rights to a given inode | |
233 | * @inode: inode to check permission on | |
234 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
235 | * | |
236 | * Used to check for read/write/execute permissions on an inode. | |
237 | * We use "fsuid" for this, letting us set arbitrary permissions | |
238 | * for filesystem access without changing the "normal" uids which | |
239 | * are used for other things. | |
240 | */ | |
f419a2e3 | 241 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 242 | { |
e6305c43 | 243 | int retval; |
1da177e4 LT |
244 | |
245 | if (mask & MAY_WRITE) { | |
22590e41 | 246 | umode_t mode = inode->i_mode; |
1da177e4 LT |
247 | |
248 | /* | |
249 | * Nobody gets write access to a read-only fs. | |
250 | */ | |
251 | if (IS_RDONLY(inode) && | |
252 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
253 | return -EROFS; | |
254 | ||
255 | /* | |
256 | * Nobody gets write access to an immutable file. | |
257 | */ | |
258 | if (IS_IMMUTABLE(inode)) | |
259 | return -EACCES; | |
260 | } | |
261 | ||
acfa4380 | 262 | if (inode->i_op->permission) |
b77b0646 | 263 | retval = inode->i_op->permission(inode, mask); |
f696a365 | 264 | else |
e6305c43 | 265 | retval = generic_permission(inode, mask, NULL); |
f696a365 | 266 | |
1da177e4 LT |
267 | if (retval) |
268 | return retval; | |
269 | ||
08ce5f16 SH |
270 | retval = devcgroup_inode_permission(inode, mask); |
271 | if (retval) | |
272 | return retval; | |
273 | ||
e6305c43 | 274 | return security_inode_permission(inode, |
f418b006 | 275 | mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND)); |
1da177e4 LT |
276 | } |
277 | ||
8c744fb8 CH |
278 | /** |
279 | * file_permission - check for additional access rights to a given file | |
280 | * @file: file to check access rights for | |
281 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
282 | * | |
283 | * Used to check for read/write/execute permissions on an already opened | |
284 | * file. | |
285 | * | |
286 | * Note: | |
287 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 288 | * be done using inode_permission(). |
8c744fb8 CH |
289 | */ |
290 | int file_permission(struct file *file, int mask) | |
291 | { | |
f419a2e3 | 292 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
293 | } |
294 | ||
1da177e4 LT |
295 | /* |
296 | * get_write_access() gets write permission for a file. | |
297 | * put_write_access() releases this write permission. | |
298 | * This is used for regular files. | |
299 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
300 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
301 | * can have the following values: | |
302 | * 0: no writers, no VM_DENYWRITE mappings | |
303 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
304 | * > 0: (i_writecount) users are writing to the file. | |
305 | * | |
306 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
307 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
308 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
309 | * to do the change if sign is wrong. Exclusion between them is provided by | |
310 | * the inode->i_lock spinlock. | |
311 | */ | |
312 | ||
313 | int get_write_access(struct inode * inode) | |
314 | { | |
315 | spin_lock(&inode->i_lock); | |
316 | if (atomic_read(&inode->i_writecount) < 0) { | |
317 | spin_unlock(&inode->i_lock); | |
318 | return -ETXTBSY; | |
319 | } | |
320 | atomic_inc(&inode->i_writecount); | |
321 | spin_unlock(&inode->i_lock); | |
322 | ||
323 | return 0; | |
324 | } | |
325 | ||
326 | int deny_write_access(struct file * file) | |
327 | { | |
0f7fc9e4 | 328 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
329 | |
330 | spin_lock(&inode->i_lock); | |
331 | if (atomic_read(&inode->i_writecount) > 0) { | |
332 | spin_unlock(&inode->i_lock); | |
333 | return -ETXTBSY; | |
334 | } | |
335 | atomic_dec(&inode->i_writecount); | |
336 | spin_unlock(&inode->i_lock); | |
337 | ||
338 | return 0; | |
339 | } | |
340 | ||
5dd784d0 JB |
341 | /** |
342 | * path_get - get a reference to a path | |
343 | * @path: path to get the reference to | |
344 | * | |
345 | * Given a path increment the reference count to the dentry and the vfsmount. | |
346 | */ | |
347 | void path_get(struct path *path) | |
348 | { | |
349 | mntget(path->mnt); | |
350 | dget(path->dentry); | |
351 | } | |
352 | EXPORT_SYMBOL(path_get); | |
353 | ||
1d957f9b JB |
354 | /** |
355 | * path_put - put a reference to a path | |
356 | * @path: path to put the reference to | |
357 | * | |
358 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
359 | */ | |
360 | void path_put(struct path *path) | |
1da177e4 | 361 | { |
1d957f9b JB |
362 | dput(path->dentry); |
363 | mntput(path->mnt); | |
1da177e4 | 364 | } |
1d957f9b | 365 | EXPORT_SYMBOL(path_put); |
1da177e4 | 366 | |
834f2a4a TM |
367 | /** |
368 | * release_open_intent - free up open intent resources | |
369 | * @nd: pointer to nameidata | |
370 | */ | |
371 | void release_open_intent(struct nameidata *nd) | |
372 | { | |
0f7fc9e4 | 373 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
374 | put_filp(nd->intent.open.file); |
375 | else | |
376 | fput(nd->intent.open.file); | |
377 | } | |
378 | ||
bcdc5e01 IK |
379 | static inline struct dentry * |
380 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
381 | { | |
382 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
383 | if (unlikely(status <= 0)) { | |
384 | /* | |
385 | * The dentry failed validation. | |
386 | * If d_revalidate returned 0 attempt to invalidate | |
387 | * the dentry otherwise d_revalidate is asking us | |
388 | * to return a fail status. | |
389 | */ | |
390 | if (!status) { | |
391 | if (!d_invalidate(dentry)) { | |
392 | dput(dentry); | |
393 | dentry = NULL; | |
394 | } | |
395 | } else { | |
396 | dput(dentry); | |
397 | dentry = ERR_PTR(status); | |
398 | } | |
399 | } | |
400 | return dentry; | |
401 | } | |
402 | ||
1da177e4 LT |
403 | /* |
404 | * Internal lookup() using the new generic dcache. | |
405 | * SMP-safe | |
406 | */ | |
407 | static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
408 | { | |
409 | struct dentry * dentry = __d_lookup(parent, name); | |
410 | ||
411 | /* lockess __d_lookup may fail due to concurrent d_move() | |
412 | * in some unrelated directory, so try with d_lookup | |
413 | */ | |
414 | if (!dentry) | |
415 | dentry = d_lookup(parent, name); | |
416 | ||
bcdc5e01 IK |
417 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) |
418 | dentry = do_revalidate(dentry, nd); | |
419 | ||
1da177e4 LT |
420 | return dentry; |
421 | } | |
422 | ||
423 | /* | |
424 | * Short-cut version of permission(), for calling by | |
425 | * path_walk(), when dcache lock is held. Combines parts | |
426 | * of permission() and generic_permission(), and tests ONLY for | |
427 | * MAY_EXEC permission. | |
428 | * | |
429 | * If appropriate, check DAC only. If not appropriate, or | |
430 | * short-cut DAC fails, then call permission() to do more | |
431 | * complete permission check. | |
432 | */ | |
672b16b2 | 433 | static int exec_permission_lite(struct inode *inode) |
1da177e4 LT |
434 | { |
435 | umode_t mode = inode->i_mode; | |
436 | ||
acfa4380 | 437 | if (inode->i_op->permission) |
1da177e4 LT |
438 | return -EAGAIN; |
439 | ||
da9592ed | 440 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
441 | mode >>= 6; |
442 | else if (in_group_p(inode->i_gid)) | |
443 | mode >>= 3; | |
444 | ||
445 | if (mode & MAY_EXEC) | |
446 | goto ok; | |
447 | ||
448 | if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE)) | |
449 | goto ok; | |
450 | ||
451 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE)) | |
452 | goto ok; | |
453 | ||
454 | if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH)) | |
455 | goto ok; | |
456 | ||
457 | return -EACCES; | |
458 | ok: | |
b77b0646 | 459 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
460 | } |
461 | ||
462 | /* | |
463 | * This is called when everything else fails, and we actually have | |
464 | * to go to the low-level filesystem to find out what we should do.. | |
465 | * | |
466 | * We get the directory semaphore, and after getting that we also | |
467 | * make sure that nobody added the entry to the dcache in the meantime.. | |
468 | * SMP-safe | |
469 | */ | |
470 | static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd) | |
471 | { | |
472 | struct dentry * result; | |
473 | struct inode *dir = parent->d_inode; | |
474 | ||
1b1dcc1b | 475 | mutex_lock(&dir->i_mutex); |
1da177e4 LT |
476 | /* |
477 | * First re-do the cached lookup just in case it was created | |
478 | * while we waited for the directory semaphore.. | |
479 | * | |
480 | * FIXME! This could use version numbering or similar to | |
481 | * avoid unnecessary cache lookups. | |
482 | * | |
483 | * The "dcache_lock" is purely to protect the RCU list walker | |
484 | * from concurrent renames at this point (we mustn't get false | |
485 | * negatives from the RCU list walk here, unlike the optimistic | |
486 | * fast walk). | |
487 | * | |
488 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
489 | */ | |
490 | result = d_lookup(parent, name); | |
491 | if (!result) { | |
d70b67c8 MS |
492 | struct dentry *dentry; |
493 | ||
494 | /* Don't create child dentry for a dead directory. */ | |
495 | result = ERR_PTR(-ENOENT); | |
496 | if (IS_DEADDIR(dir)) | |
497 | goto out_unlock; | |
498 | ||
499 | dentry = d_alloc(parent, name); | |
1da177e4 LT |
500 | result = ERR_PTR(-ENOMEM); |
501 | if (dentry) { | |
502 | result = dir->i_op->lookup(dir, dentry, nd); | |
503 | if (result) | |
504 | dput(dentry); | |
505 | else | |
506 | result = dentry; | |
507 | } | |
d70b67c8 | 508 | out_unlock: |
1b1dcc1b | 509 | mutex_unlock(&dir->i_mutex); |
1da177e4 LT |
510 | return result; |
511 | } | |
512 | ||
513 | /* | |
514 | * Uhhuh! Nasty case: the cache was re-populated while | |
515 | * we waited on the semaphore. Need to revalidate. | |
516 | */ | |
1b1dcc1b | 517 | mutex_unlock(&dir->i_mutex); |
1da177e4 | 518 | if (result->d_op && result->d_op->d_revalidate) { |
bcdc5e01 IK |
519 | result = do_revalidate(result, nd); |
520 | if (!result) | |
1da177e4 | 521 | result = ERR_PTR(-ENOENT); |
1da177e4 LT |
522 | } |
523 | return result; | |
524 | } | |
525 | ||
a02f76c3 AV |
526 | /* |
527 | * Wrapper to retry pathname resolution whenever the underlying | |
528 | * file system returns an ESTALE. | |
529 | * | |
530 | * Retry the whole path once, forcing real lookup requests | |
531 | * instead of relying on the dcache. | |
532 | */ | |
533 | static __always_inline int link_path_walk(const char *name, struct nameidata *nd) | |
534 | { | |
535 | struct path save = nd->path; | |
536 | int result; | |
537 | ||
538 | /* make sure the stuff we saved doesn't go away */ | |
c8e7f449 | 539 | path_get(&save); |
a02f76c3 AV |
540 | |
541 | result = __link_path_walk(name, nd); | |
542 | if (result == -ESTALE) { | |
543 | /* nd->path had been dropped */ | |
544 | nd->path = save; | |
c8e7f449 | 545 | path_get(&nd->path); |
a02f76c3 AV |
546 | nd->flags |= LOOKUP_REVAL; |
547 | result = __link_path_walk(name, nd); | |
548 | } | |
549 | ||
550 | path_put(&save); | |
551 | ||
552 | return result; | |
553 | } | |
554 | ||
f1662356 | 555 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
556 | { |
557 | int res = 0; | |
558 | char *name; | |
559 | if (IS_ERR(link)) | |
560 | goto fail; | |
561 | ||
562 | if (*link == '/') { | |
b4091d5f CH |
563 | struct fs_struct *fs = current->fs; |
564 | ||
1d957f9b | 565 | path_put(&nd->path); |
b4091d5f CH |
566 | |
567 | read_lock(&fs->lock); | |
568 | nd->path = fs->root; | |
569 | path_get(&fs->root); | |
570 | read_unlock(&fs->lock); | |
1da177e4 | 571 | } |
b4091d5f | 572 | |
1da177e4 | 573 | res = link_path_walk(link, nd); |
1da177e4 LT |
574 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
575 | return res; | |
576 | /* | |
577 | * If it is an iterative symlinks resolution in open_namei() we | |
578 | * have to copy the last component. And all that crap because of | |
579 | * bloody create() on broken symlinks. Furrfu... | |
580 | */ | |
581 | name = __getname(); | |
582 | if (unlikely(!name)) { | |
1d957f9b | 583 | path_put(&nd->path); |
1da177e4 LT |
584 | return -ENOMEM; |
585 | } | |
586 | strcpy(name, nd->last.name); | |
587 | nd->last.name = name; | |
588 | return 0; | |
589 | fail: | |
1d957f9b | 590 | path_put(&nd->path); |
1da177e4 LT |
591 | return PTR_ERR(link); |
592 | } | |
593 | ||
1d957f9b | 594 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
595 | { |
596 | dput(path->dentry); | |
4ac91378 | 597 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
598 | mntput(path->mnt); |
599 | } | |
600 | ||
601 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
602 | { | |
4ac91378 JB |
603 | dput(nd->path.dentry); |
604 | if (nd->path.mnt != path->mnt) | |
605 | mntput(nd->path.mnt); | |
606 | nd->path.mnt = path->mnt; | |
607 | nd->path.dentry = path->dentry; | |
051d3812 IK |
608 | } |
609 | ||
f1662356 | 610 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
611 | { |
612 | int error; | |
cc314eef | 613 | void *cookie; |
cd4e91d3 | 614 | struct dentry *dentry = path->dentry; |
1da177e4 | 615 | |
d671a1cb | 616 | touch_atime(path->mnt, dentry); |
1da177e4 | 617 | nd_set_link(nd, NULL); |
cd4e91d3 | 618 | |
4ac91378 | 619 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
620 | path_to_nameidata(path, nd); |
621 | dget(dentry); | |
622 | } | |
623 | mntget(path->mnt); | |
cc314eef LT |
624 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
625 | error = PTR_ERR(cookie); | |
626 | if (!IS_ERR(cookie)) { | |
1da177e4 | 627 | char *s = nd_get_link(nd); |
cc314eef | 628 | error = 0; |
1da177e4 LT |
629 | if (s) |
630 | error = __vfs_follow_link(nd, s); | |
631 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 632 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 633 | } |
09da5916 | 634 | path_put(path); |
1da177e4 LT |
635 | |
636 | return error; | |
637 | } | |
638 | ||
639 | /* | |
640 | * This limits recursive symlink follows to 8, while | |
641 | * limiting consecutive symlinks to 40. | |
642 | * | |
643 | * Without that kind of total limit, nasty chains of consecutive | |
644 | * symlinks can cause almost arbitrarily long lookups. | |
645 | */ | |
90ebe565 | 646 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
647 | { |
648 | int err = -ELOOP; | |
649 | if (current->link_count >= MAX_NESTED_LINKS) | |
650 | goto loop; | |
651 | if (current->total_link_count >= 40) | |
652 | goto loop; | |
653 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
654 | cond_resched(); | |
90ebe565 | 655 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
656 | if (err) |
657 | goto loop; | |
658 | current->link_count++; | |
659 | current->total_link_count++; | |
660 | nd->depth++; | |
cd4e91d3 | 661 | err = __do_follow_link(path, nd); |
839d9f93 AV |
662 | current->link_count--; |
663 | nd->depth--; | |
1da177e4 LT |
664 | return err; |
665 | loop: | |
1d957f9b JB |
666 | path_put_conditional(path, nd); |
667 | path_put(&nd->path); | |
1da177e4 LT |
668 | return err; |
669 | } | |
670 | ||
671 | int follow_up(struct vfsmount **mnt, struct dentry **dentry) | |
672 | { | |
673 | struct vfsmount *parent; | |
674 | struct dentry *mountpoint; | |
675 | spin_lock(&vfsmount_lock); | |
676 | parent=(*mnt)->mnt_parent; | |
677 | if (parent == *mnt) { | |
678 | spin_unlock(&vfsmount_lock); | |
679 | return 0; | |
680 | } | |
681 | mntget(parent); | |
682 | mountpoint=dget((*mnt)->mnt_mountpoint); | |
683 | spin_unlock(&vfsmount_lock); | |
684 | dput(*dentry); | |
685 | *dentry = mountpoint; | |
686 | mntput(*mnt); | |
687 | *mnt = parent; | |
688 | return 1; | |
689 | } | |
690 | ||
691 | /* no need for dcache_lock, as serialization is taken care in | |
692 | * namespace.c | |
693 | */ | |
463ffb2e AV |
694 | static int __follow_mount(struct path *path) |
695 | { | |
696 | int res = 0; | |
697 | while (d_mountpoint(path->dentry)) { | |
698 | struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry); | |
699 | if (!mounted) | |
700 | break; | |
701 | dput(path->dentry); | |
702 | if (res) | |
703 | mntput(path->mnt); | |
704 | path->mnt = mounted; | |
705 | path->dentry = dget(mounted->mnt_root); | |
706 | res = 1; | |
707 | } | |
708 | return res; | |
709 | } | |
710 | ||
58c465eb | 711 | static void follow_mount(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 | 712 | { |
1da177e4 LT |
713 | while (d_mountpoint(*dentry)) { |
714 | struct vfsmount *mounted = lookup_mnt(*mnt, *dentry); | |
715 | if (!mounted) | |
716 | break; | |
58c465eb | 717 | dput(*dentry); |
1da177e4 LT |
718 | mntput(*mnt); |
719 | *mnt = mounted; | |
1da177e4 | 720 | *dentry = dget(mounted->mnt_root); |
1da177e4 | 721 | } |
1da177e4 LT |
722 | } |
723 | ||
724 | /* no need for dcache_lock, as serialization is taken care in | |
725 | * namespace.c | |
726 | */ | |
e13b210f | 727 | int follow_down(struct vfsmount **mnt, struct dentry **dentry) |
1da177e4 LT |
728 | { |
729 | struct vfsmount *mounted; | |
730 | ||
731 | mounted = lookup_mnt(*mnt, *dentry); | |
732 | if (mounted) { | |
e13b210f | 733 | dput(*dentry); |
1da177e4 LT |
734 | mntput(*mnt); |
735 | *mnt = mounted; | |
1da177e4 LT |
736 | *dentry = dget(mounted->mnt_root); |
737 | return 1; | |
738 | } | |
739 | return 0; | |
740 | } | |
741 | ||
f1662356 | 742 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 743 | { |
e518ddb7 AM |
744 | struct fs_struct *fs = current->fs; |
745 | ||
1da177e4 LT |
746 | while(1) { |
747 | struct vfsmount *parent; | |
4ac91378 | 748 | struct dentry *old = nd->path.dentry; |
1da177e4 | 749 | |
e518ddb7 | 750 | read_lock(&fs->lock); |
6ac08c39 JB |
751 | if (nd->path.dentry == fs->root.dentry && |
752 | nd->path.mnt == fs->root.mnt) { | |
e518ddb7 | 753 | read_unlock(&fs->lock); |
1da177e4 LT |
754 | break; |
755 | } | |
e518ddb7 | 756 | read_unlock(&fs->lock); |
1da177e4 | 757 | spin_lock(&dcache_lock); |
4ac91378 JB |
758 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
759 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
760 | spin_unlock(&dcache_lock); |
761 | dput(old); | |
762 | break; | |
763 | } | |
764 | spin_unlock(&dcache_lock); | |
765 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
766 | parent = nd->path.mnt->mnt_parent; |
767 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
768 | spin_unlock(&vfsmount_lock); |
769 | break; | |
770 | } | |
771 | mntget(parent); | |
4ac91378 | 772 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
773 | spin_unlock(&vfsmount_lock); |
774 | dput(old); | |
4ac91378 JB |
775 | mntput(nd->path.mnt); |
776 | nd->path.mnt = parent; | |
1da177e4 | 777 | } |
4ac91378 | 778 | follow_mount(&nd->path.mnt, &nd->path.dentry); |
1da177e4 LT |
779 | } |
780 | ||
1da177e4 LT |
781 | /* |
782 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
783 | * small and for now I'd prefer to have fast path as straight as possible. | |
784 | * It _is_ time-critical. | |
785 | */ | |
786 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
787 | struct path *path) | |
788 | { | |
4ac91378 JB |
789 | struct vfsmount *mnt = nd->path.mnt; |
790 | struct dentry *dentry = __d_lookup(nd->path.dentry, name); | |
1da177e4 LT |
791 | |
792 | if (!dentry) | |
793 | goto need_lookup; | |
794 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
795 | goto need_revalidate; | |
796 | done: | |
797 | path->mnt = mnt; | |
798 | path->dentry = dentry; | |
634ee701 | 799 | __follow_mount(path); |
1da177e4 LT |
800 | return 0; |
801 | ||
802 | need_lookup: | |
4ac91378 | 803 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
804 | if (IS_ERR(dentry)) |
805 | goto fail; | |
806 | goto done; | |
807 | ||
808 | need_revalidate: | |
bcdc5e01 IK |
809 | dentry = do_revalidate(dentry, nd); |
810 | if (!dentry) | |
811 | goto need_lookup; | |
812 | if (IS_ERR(dentry)) | |
813 | goto fail; | |
814 | goto done; | |
1da177e4 LT |
815 | |
816 | fail: | |
817 | return PTR_ERR(dentry); | |
818 | } | |
819 | ||
820 | /* | |
821 | * Name resolution. | |
ea3834d9 PM |
822 | * This is the basic name resolution function, turning a pathname into |
823 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 824 | * |
ea3834d9 PM |
825 | * Returns 0 and nd will have valid dentry and mnt on success. |
826 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 827 | */ |
fc9b52cd | 828 | static int __link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
829 | { |
830 | struct path next; | |
831 | struct inode *inode; | |
832 | int err; | |
833 | unsigned int lookup_flags = nd->flags; | |
834 | ||
835 | while (*name=='/') | |
836 | name++; | |
837 | if (!*name) | |
838 | goto return_reval; | |
839 | ||
4ac91378 | 840 | inode = nd->path.dentry->d_inode; |
1da177e4 | 841 | if (nd->depth) |
f55eab82 | 842 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
843 | |
844 | /* At this point we know we have a real path component. */ | |
845 | for(;;) { | |
846 | unsigned long hash; | |
847 | struct qstr this; | |
848 | unsigned int c; | |
849 | ||
cdce5d6b | 850 | nd->flags |= LOOKUP_CONTINUE; |
672b16b2 | 851 | err = exec_permission_lite(inode); |
e4543edd | 852 | if (err == -EAGAIN) |
cb23beb5 CH |
853 | err = inode_permission(nd->path.dentry->d_inode, |
854 | MAY_EXEC); | |
6146f0d5 | 855 | if (!err) |
b9fc745d MZ |
856 | err = ima_path_check(&nd->path, MAY_EXEC, |
857 | IMA_COUNT_UPDATE); | |
1da177e4 LT |
858 | if (err) |
859 | break; | |
860 | ||
861 | this.name = name; | |
862 | c = *(const unsigned char *)name; | |
863 | ||
864 | hash = init_name_hash(); | |
865 | do { | |
866 | name++; | |
867 | hash = partial_name_hash(c, hash); | |
868 | c = *(const unsigned char *)name; | |
869 | } while (c && (c != '/')); | |
870 | this.len = name - (const char *) this.name; | |
871 | this.hash = end_name_hash(hash); | |
872 | ||
873 | /* remove trailing slashes? */ | |
874 | if (!c) | |
875 | goto last_component; | |
876 | while (*++name == '/'); | |
877 | if (!*name) | |
878 | goto last_with_slashes; | |
879 | ||
880 | /* | |
881 | * "." and ".." are special - ".." especially so because it has | |
882 | * to be able to know about the current root directory and | |
883 | * parent relationships. | |
884 | */ | |
885 | if (this.name[0] == '.') switch (this.len) { | |
886 | default: | |
887 | break; | |
888 | case 2: | |
889 | if (this.name[1] != '.') | |
890 | break; | |
58c465eb | 891 | follow_dotdot(nd); |
4ac91378 | 892 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
893 | /* fallthrough */ |
894 | case 1: | |
895 | continue; | |
896 | } | |
897 | /* | |
898 | * See if the low-level filesystem might want | |
899 | * to use its own hash.. | |
900 | */ | |
4ac91378 JB |
901 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
902 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
903 | &this); | |
1da177e4 LT |
904 | if (err < 0) |
905 | break; | |
906 | } | |
1da177e4 LT |
907 | /* This does the actual lookups.. */ |
908 | err = do_lookup(nd, &this, &next); | |
909 | if (err) | |
910 | break; | |
1da177e4 LT |
911 | |
912 | err = -ENOENT; | |
913 | inode = next.dentry->d_inode; | |
914 | if (!inode) | |
915 | goto out_dput; | |
1da177e4 LT |
916 | |
917 | if (inode->i_op->follow_link) { | |
90ebe565 | 918 | err = do_follow_link(&next, nd); |
1da177e4 LT |
919 | if (err) |
920 | goto return_err; | |
921 | err = -ENOENT; | |
4ac91378 | 922 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
923 | if (!inode) |
924 | break; | |
09dd17d3 MS |
925 | } else |
926 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
927 | err = -ENOTDIR; |
928 | if (!inode->i_op->lookup) | |
929 | break; | |
930 | continue; | |
931 | /* here ends the main loop */ | |
932 | ||
933 | last_with_slashes: | |
934 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
935 | last_component: | |
f55eab82 TM |
936 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
937 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
938 | if (lookup_flags & LOOKUP_PARENT) |
939 | goto lookup_parent; | |
940 | if (this.name[0] == '.') switch (this.len) { | |
941 | default: | |
942 | break; | |
943 | case 2: | |
944 | if (this.name[1] != '.') | |
945 | break; | |
58c465eb | 946 | follow_dotdot(nd); |
4ac91378 | 947 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
948 | /* fallthrough */ |
949 | case 1: | |
950 | goto return_reval; | |
951 | } | |
4ac91378 JB |
952 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
953 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
954 | &this); | |
1da177e4 LT |
955 | if (err < 0) |
956 | break; | |
957 | } | |
958 | err = do_lookup(nd, &this, &next); | |
959 | if (err) | |
960 | break; | |
1da177e4 LT |
961 | inode = next.dentry->d_inode; |
962 | if ((lookup_flags & LOOKUP_FOLLOW) | |
acfa4380 | 963 | && inode && inode->i_op->follow_link) { |
90ebe565 | 964 | err = do_follow_link(&next, nd); |
1da177e4 LT |
965 | if (err) |
966 | goto return_err; | |
4ac91378 | 967 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
968 | } else |
969 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
970 | err = -ENOENT; |
971 | if (!inode) | |
972 | break; | |
973 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
974 | err = -ENOTDIR; | |
acfa4380 | 975 | if (!inode->i_op->lookup) |
1da177e4 LT |
976 | break; |
977 | } | |
978 | goto return_base; | |
979 | lookup_parent: | |
980 | nd->last = this; | |
981 | nd->last_type = LAST_NORM; | |
982 | if (this.name[0] != '.') | |
983 | goto return_base; | |
984 | if (this.len == 1) | |
985 | nd->last_type = LAST_DOT; | |
986 | else if (this.len == 2 && this.name[1] == '.') | |
987 | nd->last_type = LAST_DOTDOT; | |
988 | else | |
989 | goto return_base; | |
990 | return_reval: | |
991 | /* | |
992 | * We bypassed the ordinary revalidation routines. | |
993 | * We may need to check the cached dentry for staleness. | |
994 | */ | |
4ac91378 JB |
995 | if (nd->path.dentry && nd->path.dentry->d_sb && |
996 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
997 | err = -ESTALE; |
998 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
999 | if (!nd->path.dentry->d_op->d_revalidate( |
1000 | nd->path.dentry, nd)) | |
1da177e4 LT |
1001 | break; |
1002 | } | |
1003 | return_base: | |
1004 | return 0; | |
1005 | out_dput: | |
1d957f9b | 1006 | path_put_conditional(&next, nd); |
1da177e4 LT |
1007 | break; |
1008 | } | |
1d957f9b | 1009 | path_put(&nd->path); |
1da177e4 LT |
1010 | return_err: |
1011 | return err; | |
1012 | } | |
1013 | ||
fc9b52cd | 1014 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1015 | { |
1016 | current->total_link_count = 0; | |
1017 | return link_path_walk(name, nd); | |
1018 | } | |
1019 | ||
9b4a9b14 | 1020 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1021 | { |
ea3834d9 | 1022 | int retval = 0; |
170aa3d0 UD |
1023 | int fput_needed; |
1024 | struct file *file; | |
e518ddb7 | 1025 | struct fs_struct *fs = current->fs; |
1da177e4 LT |
1026 | |
1027 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1028 | nd->flags = flags; | |
1029 | nd->depth = 0; | |
1030 | ||
1da177e4 | 1031 | if (*name=='/') { |
e518ddb7 | 1032 | read_lock(&fs->lock); |
6ac08c39 JB |
1033 | nd->path = fs->root; |
1034 | path_get(&fs->root); | |
e518ddb7 | 1035 | read_unlock(&fs->lock); |
5590ff0d | 1036 | } else if (dfd == AT_FDCWD) { |
e518ddb7 | 1037 | read_lock(&fs->lock); |
6ac08c39 JB |
1038 | nd->path = fs->pwd; |
1039 | path_get(&fs->pwd); | |
e518ddb7 | 1040 | read_unlock(&fs->lock); |
5590ff0d | 1041 | } else { |
5590ff0d UD |
1042 | struct dentry *dentry; |
1043 | ||
1044 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1045 | retval = -EBADF; |
1046 | if (!file) | |
6d09bb62 | 1047 | goto out_fail; |
5590ff0d | 1048 | |
0f7fc9e4 | 1049 | dentry = file->f_path.dentry; |
5590ff0d | 1050 | |
170aa3d0 UD |
1051 | retval = -ENOTDIR; |
1052 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1053 | goto fput_fail; |
5590ff0d UD |
1054 | |
1055 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1056 | if (retval) |
6d09bb62 | 1057 | goto fput_fail; |
5590ff0d | 1058 | |
5dd784d0 JB |
1059 | nd->path = file->f_path; |
1060 | path_get(&file->f_path); | |
5590ff0d UD |
1061 | |
1062 | fput_light(file, fput_needed); | |
1da177e4 | 1063 | } |
9b4a9b14 | 1064 | return 0; |
2dfdd266 | 1065 | |
9b4a9b14 AV |
1066 | fput_fail: |
1067 | fput_light(file, fput_needed); | |
1068 | out_fail: | |
1069 | return retval; | |
1070 | } | |
1071 | ||
1072 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1073 | static int do_path_lookup(int dfd, const char *name, | |
1074 | unsigned int flags, struct nameidata *nd) | |
1075 | { | |
1076 | int retval = path_init(dfd, name, flags, nd); | |
1077 | if (!retval) | |
1078 | retval = path_walk(name, nd); | |
4ac91378 JB |
1079 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1080 | nd->path.dentry->d_inode)) | |
1081 | audit_inode(name, nd->path.dentry); | |
170aa3d0 | 1082 | return retval; |
1da177e4 LT |
1083 | } |
1084 | ||
fc9b52cd | 1085 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1086 | struct nameidata *nd) |
1087 | { | |
1088 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1089 | } | |
1090 | ||
d1811465 AV |
1091 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1092 | { | |
1093 | struct nameidata nd; | |
1094 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1095 | if (!res) | |
1096 | *path = nd.path; | |
1097 | return res; | |
1098 | } | |
1099 | ||
16f18200 JJS |
1100 | /** |
1101 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1102 | * @dentry: pointer to dentry of the base directory | |
1103 | * @mnt: pointer to vfs mount of the base directory | |
1104 | * @name: pointer to file name | |
1105 | * @flags: lookup flags | |
1106 | * @nd: pointer to nameidata | |
1107 | */ | |
1108 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1109 | const char *name, unsigned int flags, | |
1110 | struct nameidata *nd) | |
1111 | { | |
1112 | int retval; | |
1113 | ||
1114 | /* same as do_path_lookup */ | |
1115 | nd->last_type = LAST_ROOT; | |
1116 | nd->flags = flags; | |
1117 | nd->depth = 0; | |
1118 | ||
c8e7f449 JB |
1119 | nd->path.dentry = dentry; |
1120 | nd->path.mnt = mnt; | |
1121 | path_get(&nd->path); | |
16f18200 JJS |
1122 | |
1123 | retval = path_walk(name, nd); | |
4ac91378 JB |
1124 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1125 | nd->path.dentry->d_inode)) | |
1126 | audit_inode(name, nd->path.dentry); | |
16f18200 JJS |
1127 | |
1128 | return retval; | |
1129 | ||
1130 | } | |
1131 | ||
8737f3a1 AV |
1132 | /** |
1133 | * path_lookup_open - lookup a file path with open intent | |
1134 | * @dfd: the directory to use as base, or AT_FDCWD | |
1135 | * @name: pointer to file name | |
1136 | * @lookup_flags: lookup intent flags | |
1137 | * @nd: pointer to nameidata | |
1138 | * @open_flags: open intent flags | |
1139 | */ | |
6e8341a1 AV |
1140 | static int path_lookup_open(int dfd, const char *name, |
1141 | unsigned int lookup_flags, struct nameidata *nd, int open_flags) | |
834f2a4a TM |
1142 | { |
1143 | struct file *filp = get_empty_filp(); | |
1144 | int err; | |
1145 | ||
1146 | if (filp == NULL) | |
1147 | return -ENFILE; | |
1148 | nd->intent.open.file = filp; | |
1149 | nd->intent.open.flags = open_flags; | |
8737f3a1 | 1150 | nd->intent.open.create_mode = 0; |
5590ff0d | 1151 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1152 | if (IS_ERR(nd->intent.open.file)) { |
1153 | if (err == 0) { | |
1154 | err = PTR_ERR(nd->intent.open.file); | |
1d957f9b | 1155 | path_put(&nd->path); |
834f2a4a TM |
1156 | } |
1157 | } else if (err != 0) | |
1158 | release_open_intent(nd); | |
1159 | return err; | |
1160 | } | |
1161 | ||
eead1911 CH |
1162 | static struct dentry *__lookup_hash(struct qstr *name, |
1163 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1164 | { |
057f6c01 | 1165 | struct dentry *dentry; |
1da177e4 LT |
1166 | struct inode *inode; |
1167 | int err; | |
1168 | ||
1169 | inode = base->d_inode; | |
1da177e4 LT |
1170 | |
1171 | /* | |
1172 | * See if the low-level filesystem might want | |
1173 | * to use its own hash.. | |
1174 | */ | |
1175 | if (base->d_op && base->d_op->d_hash) { | |
1176 | err = base->d_op->d_hash(base, name); | |
1177 | dentry = ERR_PTR(err); | |
1178 | if (err < 0) | |
1179 | goto out; | |
1180 | } | |
1181 | ||
1182 | dentry = cached_lookup(base, name, nd); | |
1183 | if (!dentry) { | |
d70b67c8 MS |
1184 | struct dentry *new; |
1185 | ||
1186 | /* Don't create child dentry for a dead directory. */ | |
1187 | dentry = ERR_PTR(-ENOENT); | |
1188 | if (IS_DEADDIR(inode)) | |
1189 | goto out; | |
1190 | ||
1191 | new = d_alloc(base, name); | |
1da177e4 LT |
1192 | dentry = ERR_PTR(-ENOMEM); |
1193 | if (!new) | |
1194 | goto out; | |
1195 | dentry = inode->i_op->lookup(inode, new, nd); | |
1196 | if (!dentry) | |
1197 | dentry = new; | |
1198 | else | |
1199 | dput(new); | |
1200 | } | |
1201 | out: | |
1202 | return dentry; | |
1203 | } | |
1204 | ||
057f6c01 JM |
1205 | /* |
1206 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1207 | * needs parent already locked. Doesn't follow mounts. | |
1208 | * SMP-safe. | |
1209 | */ | |
eead1911 | 1210 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1211 | { |
057f6c01 JM |
1212 | int err; |
1213 | ||
f419a2e3 | 1214 | err = inode_permission(nd->path.dentry->d_inode, MAY_EXEC); |
057f6c01 | 1215 | if (err) |
eead1911 | 1216 | return ERR_PTR(err); |
4ac91378 | 1217 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1218 | } |
1219 | ||
eead1911 CH |
1220 | static int __lookup_one_len(const char *name, struct qstr *this, |
1221 | struct dentry *base, int len) | |
1da177e4 LT |
1222 | { |
1223 | unsigned long hash; | |
1da177e4 LT |
1224 | unsigned int c; |
1225 | ||
057f6c01 JM |
1226 | this->name = name; |
1227 | this->len = len; | |
1da177e4 | 1228 | if (!len) |
057f6c01 | 1229 | return -EACCES; |
1da177e4 LT |
1230 | |
1231 | hash = init_name_hash(); | |
1232 | while (len--) { | |
1233 | c = *(const unsigned char *)name++; | |
1234 | if (c == '/' || c == '\0') | |
057f6c01 | 1235 | return -EACCES; |
1da177e4 LT |
1236 | hash = partial_name_hash(c, hash); |
1237 | } | |
057f6c01 JM |
1238 | this->hash = end_name_hash(hash); |
1239 | return 0; | |
1240 | } | |
1da177e4 | 1241 | |
eead1911 | 1242 | /** |
a6b91919 | 1243 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1244 | * @name: pathname component to lookup |
1245 | * @base: base directory to lookup from | |
1246 | * @len: maximum length @len should be interpreted to | |
1247 | * | |
a6b91919 RD |
1248 | * Note that this routine is purely a helper for filesystem usage and should |
1249 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1250 | * nameidata argument is passed to the filesystem methods and a filesystem |
1251 | * using this helper needs to be prepared for that. | |
1252 | */ | |
057f6c01 JM |
1253 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1254 | { | |
1255 | int err; | |
1256 | struct qstr this; | |
1257 | ||
2f9092e1 DW |
1258 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1259 | ||
057f6c01 | 1260 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1261 | if (err) |
1262 | return ERR_PTR(err); | |
1263 | ||
f419a2e3 | 1264 | err = inode_permission(base->d_inode, MAY_EXEC); |
057f6c01 JM |
1265 | if (err) |
1266 | return ERR_PTR(err); | |
49705b77 | 1267 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1268 | } |
1269 | ||
eead1911 CH |
1270 | /** |
1271 | * lookup_one_noperm - bad hack for sysfs | |
1272 | * @name: pathname component to lookup | |
1273 | * @base: base directory to lookup from | |
1274 | * | |
1275 | * This is a variant of lookup_one_len that doesn't perform any permission | |
1276 | * checks. It's a horrible hack to work around the braindead sysfs | |
1277 | * architecture and should not be used anywhere else. | |
1278 | * | |
1279 | * DON'T USE THIS FUNCTION EVER, thanks. | |
1280 | */ | |
1281 | struct dentry *lookup_one_noperm(const char *name, struct dentry *base) | |
057f6c01 JM |
1282 | { |
1283 | int err; | |
1284 | struct qstr this; | |
1285 | ||
eead1911 | 1286 | err = __lookup_one_len(name, &this, base, strlen(name)); |
057f6c01 JM |
1287 | if (err) |
1288 | return ERR_PTR(err); | |
eead1911 | 1289 | return __lookup_hash(&this, base, NULL); |
1da177e4 LT |
1290 | } |
1291 | ||
2d8f3038 AV |
1292 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1293 | struct path *path) | |
1da177e4 | 1294 | { |
2d8f3038 | 1295 | struct nameidata nd; |
1da177e4 LT |
1296 | char *tmp = getname(name); |
1297 | int err = PTR_ERR(tmp); | |
1da177e4 | 1298 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1299 | |
1300 | BUG_ON(flags & LOOKUP_PARENT); | |
1301 | ||
1302 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1303 | putname(tmp); |
2d8f3038 AV |
1304 | if (!err) |
1305 | *path = nd.path; | |
1da177e4 LT |
1306 | } |
1307 | return err; | |
1308 | } | |
1309 | ||
2ad94ae6 AV |
1310 | static int user_path_parent(int dfd, const char __user *path, |
1311 | struct nameidata *nd, char **name) | |
1312 | { | |
1313 | char *s = getname(path); | |
1314 | int error; | |
1315 | ||
1316 | if (IS_ERR(s)) | |
1317 | return PTR_ERR(s); | |
1318 | ||
1319 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1320 | if (error) | |
1321 | putname(s); | |
1322 | else | |
1323 | *name = s; | |
1324 | ||
1325 | return error; | |
1326 | } | |
1327 | ||
1da177e4 LT |
1328 | /* |
1329 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1330 | * minimal. | |
1331 | */ | |
1332 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1333 | { | |
da9592ed DH |
1334 | uid_t fsuid = current_fsuid(); |
1335 | ||
1da177e4 LT |
1336 | if (!(dir->i_mode & S_ISVTX)) |
1337 | return 0; | |
da9592ed | 1338 | if (inode->i_uid == fsuid) |
1da177e4 | 1339 | return 0; |
da9592ed | 1340 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1341 | return 0; |
1342 | return !capable(CAP_FOWNER); | |
1343 | } | |
1344 | ||
1345 | /* | |
1346 | * Check whether we can remove a link victim from directory dir, check | |
1347 | * whether the type of victim is right. | |
1348 | * 1. We can't do it if dir is read-only (done in permission()) | |
1349 | * 2. We should have write and exec permissions on dir | |
1350 | * 3. We can't remove anything from append-only dir | |
1351 | * 4. We can't do anything with immutable dir (done in permission()) | |
1352 | * 5. If the sticky bit on dir is set we should either | |
1353 | * a. be owner of dir, or | |
1354 | * b. be owner of victim, or | |
1355 | * c. have CAP_FOWNER capability | |
1356 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1357 | * links pointing to it. | |
1358 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1359 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1360 | * 9. We can't remove a root or mountpoint. | |
1361 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1362 | * nfs_async_unlink(). | |
1363 | */ | |
858119e1 | 1364 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1365 | { |
1366 | int error; | |
1367 | ||
1368 | if (!victim->d_inode) | |
1369 | return -ENOENT; | |
1370 | ||
1371 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1372 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1373 | |
f419a2e3 | 1374 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1375 | if (error) |
1376 | return error; | |
1377 | if (IS_APPEND(dir)) | |
1378 | return -EPERM; | |
1379 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1380 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1381 | return -EPERM; |
1382 | if (isdir) { | |
1383 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1384 | return -ENOTDIR; | |
1385 | if (IS_ROOT(victim)) | |
1386 | return -EBUSY; | |
1387 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1388 | return -EISDIR; | |
1389 | if (IS_DEADDIR(dir)) | |
1390 | return -ENOENT; | |
1391 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1392 | return -EBUSY; | |
1393 | return 0; | |
1394 | } | |
1395 | ||
1396 | /* Check whether we can create an object with dentry child in directory | |
1397 | * dir. | |
1398 | * 1. We can't do it if child already exists (open has special treatment for | |
1399 | * this case, but since we are inlined it's OK) | |
1400 | * 2. We can't do it if dir is read-only (done in permission()) | |
1401 | * 3. We should have write and exec permissions on dir | |
1402 | * 4. We can't do it if dir is immutable (done in permission()) | |
1403 | */ | |
a95164d9 | 1404 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1405 | { |
1406 | if (child->d_inode) | |
1407 | return -EEXIST; | |
1408 | if (IS_DEADDIR(dir)) | |
1409 | return -ENOENT; | |
f419a2e3 | 1410 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1411 | } |
1412 | ||
1413 | /* | |
1da177e4 LT |
1414 | * O_DIRECTORY translates into forcing a directory lookup. |
1415 | */ | |
1416 | static inline int lookup_flags(unsigned int f) | |
1417 | { | |
1418 | unsigned long retval = LOOKUP_FOLLOW; | |
1419 | ||
1420 | if (f & O_NOFOLLOW) | |
1421 | retval &= ~LOOKUP_FOLLOW; | |
1422 | ||
1da177e4 LT |
1423 | if (f & O_DIRECTORY) |
1424 | retval |= LOOKUP_DIRECTORY; | |
1425 | ||
1426 | return retval; | |
1427 | } | |
1428 | ||
1429 | /* | |
1430 | * p1 and p2 should be directories on the same fs. | |
1431 | */ | |
1432 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1433 | { | |
1434 | struct dentry *p; | |
1435 | ||
1436 | if (p1 == p2) { | |
f2eace23 | 1437 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1438 | return NULL; |
1439 | } | |
1440 | ||
a11f3a05 | 1441 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1442 | |
e2761a11 OH |
1443 | p = d_ancestor(p2, p1); |
1444 | if (p) { | |
1445 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1446 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1447 | return p; | |
1da177e4 LT |
1448 | } |
1449 | ||
e2761a11 OH |
1450 | p = d_ancestor(p1, p2); |
1451 | if (p) { | |
1452 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1453 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1454 | return p; | |
1da177e4 LT |
1455 | } |
1456 | ||
f2eace23 IM |
1457 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1458 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1459 | return NULL; |
1460 | } | |
1461 | ||
1462 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1463 | { | |
1b1dcc1b | 1464 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1465 | if (p1 != p2) { |
1b1dcc1b | 1466 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1467 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1468 | } |
1469 | } | |
1470 | ||
1471 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1472 | struct nameidata *nd) | |
1473 | { | |
a95164d9 | 1474 | int error = may_create(dir, dentry); |
1da177e4 LT |
1475 | |
1476 | if (error) | |
1477 | return error; | |
1478 | ||
acfa4380 | 1479 | if (!dir->i_op->create) |
1da177e4 LT |
1480 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1481 | mode &= S_IALLUGO; | |
1482 | mode |= S_IFREG; | |
1483 | error = security_inode_create(dir, dentry, mode); | |
1484 | if (error) | |
1485 | return error; | |
9e3509e2 | 1486 | vfs_dq_init(dir); |
1da177e4 | 1487 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1488 | if (!error) |
f38aa942 | 1489 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1490 | return error; |
1491 | } | |
1492 | ||
3fb64190 | 1493 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1494 | { |
3fb64190 | 1495 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1496 | struct inode *inode = dentry->d_inode; |
1497 | int error; | |
1498 | ||
1499 | if (!inode) | |
1500 | return -ENOENT; | |
1501 | ||
c8fe8f30 CH |
1502 | switch (inode->i_mode & S_IFMT) { |
1503 | case S_IFLNK: | |
1da177e4 | 1504 | return -ELOOP; |
c8fe8f30 CH |
1505 | case S_IFDIR: |
1506 | if (acc_mode & MAY_WRITE) | |
1507 | return -EISDIR; | |
1508 | break; | |
1509 | case S_IFBLK: | |
1510 | case S_IFCHR: | |
3fb64190 | 1511 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1512 | return -EACCES; |
c8fe8f30 CH |
1513 | /*FALLTHRU*/ |
1514 | case S_IFIFO: | |
1515 | case S_IFSOCK: | |
1da177e4 | 1516 | flag &= ~O_TRUNC; |
c8fe8f30 | 1517 | break; |
4a3fd211 | 1518 | } |
b41572e9 | 1519 | |
3fb64190 | 1520 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1521 | if (error) |
1522 | return error; | |
6146f0d5 | 1523 | |
cb5629b1 | 1524 | error = ima_path_check(path, |
b9fc745d MZ |
1525 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC), |
1526 | IMA_COUNT_UPDATE); | |
6146f0d5 MZ |
1527 | if (error) |
1528 | return error; | |
1da177e4 LT |
1529 | /* |
1530 | * An append-only file must be opened in append mode for writing. | |
1531 | */ | |
1532 | if (IS_APPEND(inode)) { | |
1533 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
1534 | return -EPERM; | |
1535 | if (flag & O_TRUNC) | |
1536 | return -EPERM; | |
1537 | } | |
1538 | ||
1539 | /* O_NOATIME can only be set by the owner or superuser */ | |
1540 | if (flag & O_NOATIME) | |
3bd858ab | 1541 | if (!is_owner_or_cap(inode)) |
1da177e4 LT |
1542 | return -EPERM; |
1543 | ||
1544 | /* | |
1545 | * Ensure there are no outstanding leases on the file. | |
1546 | */ | |
1547 | error = break_lease(inode, flag); | |
1548 | if (error) | |
1549 | return error; | |
1550 | ||
1551 | if (flag & O_TRUNC) { | |
1552 | error = get_write_access(inode); | |
1553 | if (error) | |
1554 | return error; | |
1555 | ||
1556 | /* | |
1557 | * Refuse to truncate files with mandatory locks held on them. | |
1558 | */ | |
1559 | error = locks_verify_locked(inode); | |
be6d3e56 | 1560 | if (!error) |
3fb64190 | 1561 | error = security_path_truncate(path, 0, |
be6d3e56 | 1562 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN); |
1da177e4 | 1563 | if (!error) { |
9e3509e2 | 1564 | vfs_dq_init(inode); |
d139d7ff MS |
1565 | |
1566 | error = do_truncate(dentry, 0, | |
1567 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1568 | NULL); | |
1da177e4 LT |
1569 | } |
1570 | put_write_access(inode); | |
1571 | if (error) | |
1572 | return error; | |
1573 | } else | |
1574 | if (flag & FMODE_WRITE) | |
9e3509e2 | 1575 | vfs_dq_init(inode); |
1da177e4 LT |
1576 | |
1577 | return 0; | |
1578 | } | |
1579 | ||
d57999e1 DH |
1580 | /* |
1581 | * Be careful about ever adding any more callers of this | |
1582 | * function. Its flags must be in the namei format, not | |
1583 | * what get passed to sys_open(). | |
1584 | */ | |
1585 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1586 | int flag, int mode) |
1587 | { | |
1588 | int error; | |
4ac91378 | 1589 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1590 | |
1591 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1592 | mode &= ~current_umask(); |
be6d3e56 KT |
1593 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1594 | if (error) | |
1595 | goto out_unlock; | |
aab520e2 | 1596 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1597 | out_unlock: |
aab520e2 | 1598 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1599 | dput(nd->path.dentry); |
1600 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1601 | if (error) |
1602 | return error; | |
1603 | /* Don't check for write permission, don't truncate */ | |
3fb64190 | 1604 | return may_open(&nd->path, 0, flag & ~O_TRUNC); |
aab520e2 DH |
1605 | } |
1606 | ||
d57999e1 DH |
1607 | /* |
1608 | * Note that while the flag value (low two bits) for sys_open means: | |
1609 | * 00 - read-only | |
1610 | * 01 - write-only | |
1611 | * 10 - read-write | |
1612 | * 11 - special | |
1613 | * it is changed into | |
1614 | * 00 - no permissions needed | |
1615 | * 01 - read-permission | |
1616 | * 10 - write-permission | |
1617 | * 11 - read-write | |
1618 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1619 | * This is more logical, and also allows the 00 "no perm needed" | |
1620 | * to be used for symlinks (where the permissions are checked | |
1621 | * later). | |
1622 | * | |
1623 | */ | |
1624 | static inline int open_to_namei_flags(int flag) | |
1625 | { | |
1626 | if ((flag+1) & O_ACCMODE) | |
1627 | flag++; | |
1628 | return flag; | |
1629 | } | |
1630 | ||
4a3fd211 DH |
1631 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1632 | { | |
1633 | /* | |
1634 | * We'll never write to the fs underlying | |
1635 | * a device file. | |
1636 | */ | |
1637 | if (special_file(inode->i_mode)) | |
1638 | return 0; | |
1639 | return (flag & O_TRUNC); | |
1640 | } | |
1641 | ||
1da177e4 | 1642 | /* |
4a3fd211 DH |
1643 | * Note that the low bits of the passed in "open_flag" |
1644 | * are not the same as in the local variable "flag". See | |
1645 | * open_to_namei_flags() for more details. | |
1da177e4 | 1646 | */ |
a70e65df | 1647 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1648 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1649 | { |
4a3fd211 | 1650 | struct file *filp; |
a70e65df | 1651 | struct nameidata nd; |
6e8341a1 | 1652 | int error; |
4e7506e4 | 1653 | struct path path; |
1da177e4 LT |
1654 | struct dentry *dir; |
1655 | int count = 0; | |
4a3fd211 | 1656 | int will_write; |
d57999e1 | 1657 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1658 | |
6e8341a1 AV |
1659 | if (!acc_mode) |
1660 | acc_mode = MAY_OPEN | ACC_MODE(flag); | |
1da177e4 | 1661 | |
834f2a4a TM |
1662 | /* O_TRUNC implies we need access checks for write permissions */ |
1663 | if (flag & O_TRUNC) | |
1664 | acc_mode |= MAY_WRITE; | |
1665 | ||
1da177e4 LT |
1666 | /* Allow the LSM permission hook to distinguish append |
1667 | access from general write access. */ | |
1668 | if (flag & O_APPEND) | |
1669 | acc_mode |= MAY_APPEND; | |
1670 | ||
1da177e4 LT |
1671 | /* |
1672 | * The simplest case - just a plain lookup. | |
1673 | */ | |
1674 | if (!(flag & O_CREAT)) { | |
5590ff0d | 1675 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
a70e65df | 1676 | &nd, flag); |
1da177e4 | 1677 | if (error) |
a70e65df | 1678 | return ERR_PTR(error); |
1da177e4 LT |
1679 | goto ok; |
1680 | } | |
1681 | ||
1682 | /* | |
1683 | * Create - we need to know the parent. | |
1684 | */ | |
9b4a9b14 | 1685 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1686 | if (error) |
a70e65df | 1687 | return ERR_PTR(error); |
9b4a9b14 AV |
1688 | error = path_walk(pathname, &nd); |
1689 | if (error) | |
1690 | return ERR_PTR(error); | |
1691 | if (unlikely(!audit_dummy_context())) | |
1692 | audit_inode(pathname, nd.path.dentry); | |
1da177e4 LT |
1693 | |
1694 | /* | |
1695 | * We have the parent and last component. First of all, check | |
1696 | * that we are not asked to creat(2) an obvious directory - that | |
1697 | * will not do. | |
1698 | */ | |
1699 | error = -EISDIR; | |
a70e65df | 1700 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1701 | goto exit_parent; |
1da177e4 | 1702 | |
8737f3a1 AV |
1703 | error = -ENFILE; |
1704 | filp = get_empty_filp(); | |
1705 | if (filp == NULL) | |
1706 | goto exit_parent; | |
1707 | nd.intent.open.file = filp; | |
1708 | nd.intent.open.flags = flag; | |
1709 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1710 | dir = nd.path.dentry; |
1711 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1712 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1713 | if (flag & O_EXCL) |
1714 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1715 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1716 | path.dentry = lookup_hash(&nd); |
1717 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1718 | |
1719 | do_last: | |
4e7506e4 AV |
1720 | error = PTR_ERR(path.dentry); |
1721 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1722 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1723 | goto exit; |
1724 | } | |
1725 | ||
a70e65df | 1726 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1727 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1728 | goto exit_mutex_unlock; |
4af4c52f OD |
1729 | } |
1730 | ||
1da177e4 | 1731 | /* Negative dentry, just create the file */ |
4e7506e4 | 1732 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1733 | /* |
1734 | * This write is needed to ensure that a | |
1735 | * ro->rw transition does not occur between | |
1736 | * the time when the file is created and when | |
1737 | * a permanent write count is taken through | |
1738 | * the 'struct file' in nameidata_to_filp(). | |
1739 | */ | |
1740 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1741 | if (error) |
4a3fd211 DH |
1742 | goto exit_mutex_unlock; |
1743 | error = __open_namei_create(&nd, &path, flag, mode); | |
1744 | if (error) { | |
1745 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1746 | goto exit; |
4a3fd211 DH |
1747 | } |
1748 | filp = nameidata_to_filp(&nd, open_flag); | |
1749 | mnt_drop_write(nd.path.mnt); | |
1750 | return filp; | |
1da177e4 LT |
1751 | } |
1752 | ||
1753 | /* | |
1754 | * It already exists. | |
1755 | */ | |
1b1dcc1b | 1756 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1757 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1758 | |
1759 | error = -EEXIST; | |
1760 | if (flag & O_EXCL) | |
1761 | goto exit_dput; | |
1762 | ||
e13b210f | 1763 | if (__follow_mount(&path)) { |
1da177e4 | 1764 | error = -ELOOP; |
ba7a4c1a AV |
1765 | if (flag & O_NOFOLLOW) |
1766 | goto exit_dput; | |
1da177e4 | 1767 | } |
3e2efce0 | 1768 | |
1da177e4 | 1769 | error = -ENOENT; |
4e7506e4 | 1770 | if (!path.dentry->d_inode) |
1da177e4 | 1771 | goto exit_dput; |
acfa4380 | 1772 | if (path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1773 | goto do_link; |
1774 | ||
a70e65df | 1775 | path_to_nameidata(&path, &nd); |
1da177e4 | 1776 | error = -EISDIR; |
4e7506e4 | 1777 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1778 | goto exit; |
1779 | ok: | |
4a3fd211 DH |
1780 | /* |
1781 | * Consider: | |
1782 | * 1. may_open() truncates a file | |
1783 | * 2. a rw->ro mount transition occurs | |
1784 | * 3. nameidata_to_filp() fails due to | |
1785 | * the ro mount. | |
1786 | * That would be inconsistent, and should | |
1787 | * be avoided. Taking this mnt write here | |
1788 | * ensures that (2) can not occur. | |
1789 | */ | |
1790 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1791 | if (will_write) { | |
1792 | error = mnt_want_write(nd.path.mnt); | |
1793 | if (error) | |
1794 | goto exit; | |
1795 | } | |
3fb64190 | 1796 | error = may_open(&nd.path, acc_mode, flag); |
4a3fd211 DH |
1797 | if (error) { |
1798 | if (will_write) | |
1799 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1800 | goto exit; |
4a3fd211 DH |
1801 | } |
1802 | filp = nameidata_to_filp(&nd, open_flag); | |
1803 | /* | |
1804 | * It is now safe to drop the mnt write | |
1805 | * because the filp has had a write taken | |
1806 | * on its behalf. | |
1807 | */ | |
1808 | if (will_write) | |
1809 | mnt_drop_write(nd.path.mnt); | |
1810 | return filp; | |
1da177e4 | 1811 | |
4a3fd211 DH |
1812 | exit_mutex_unlock: |
1813 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1814 | exit_dput: |
a70e65df | 1815 | path_put_conditional(&path, &nd); |
1da177e4 | 1816 | exit: |
a70e65df CH |
1817 | if (!IS_ERR(nd.intent.open.file)) |
1818 | release_open_intent(&nd); | |
8737f3a1 | 1819 | exit_parent: |
a70e65df CH |
1820 | path_put(&nd.path); |
1821 | return ERR_PTR(error); | |
1da177e4 LT |
1822 | |
1823 | do_link: | |
1824 | error = -ELOOP; | |
1825 | if (flag & O_NOFOLLOW) | |
1826 | goto exit_dput; | |
1827 | /* | |
1828 | * This is subtle. Instead of calling do_follow_link() we do the | |
1829 | * thing by hands. The reason is that this way we have zero link_count | |
1830 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1831 | * After that we have the parent and last component, i.e. | |
1832 | * we are in the same situation as after the first path_walk(). | |
1833 | * Well, almost - if the last component is normal we get its copy | |
1834 | * stored in nd->last.name and we will have to putname() it when we | |
1835 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1836 | */ | |
a70e65df CH |
1837 | nd.flags |= LOOKUP_PARENT; |
1838 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1839 | if (error) |
1840 | goto exit_dput; | |
a70e65df | 1841 | error = __do_follow_link(&path, &nd); |
de459215 KK |
1842 | if (error) { |
1843 | /* Does someone understand code flow here? Or it is only | |
1844 | * me so stupid? Anathema to whoever designed this non-sense | |
1845 | * with "intent.open". | |
1846 | */ | |
a70e65df CH |
1847 | release_open_intent(&nd); |
1848 | return ERR_PTR(error); | |
de459215 | 1849 | } |
a70e65df CH |
1850 | nd.flags &= ~LOOKUP_PARENT; |
1851 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1852 | goto ok; |
1da177e4 | 1853 | error = -EISDIR; |
a70e65df | 1854 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1855 | goto exit; |
a70e65df CH |
1856 | if (nd.last.name[nd.last.len]) { |
1857 | __putname(nd.last.name); | |
1da177e4 LT |
1858 | goto exit; |
1859 | } | |
1860 | error = -ELOOP; | |
1861 | if (count++==32) { | |
a70e65df | 1862 | __putname(nd.last.name); |
1da177e4 LT |
1863 | goto exit; |
1864 | } | |
a70e65df | 1865 | dir = nd.path.dentry; |
1b1dcc1b | 1866 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1867 | path.dentry = lookup_hash(&nd); |
1868 | path.mnt = nd.path.mnt; | |
1869 | __putname(nd.last.name); | |
1da177e4 LT |
1870 | goto do_last; |
1871 | } | |
1872 | ||
a70e65df CH |
1873 | /** |
1874 | * filp_open - open file and return file pointer | |
1875 | * | |
1876 | * @filename: path to open | |
1877 | * @flags: open flags as per the open(2) second argument | |
1878 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1879 | * | |
1880 | * This is the helper to open a file from kernelspace if you really | |
1881 | * have to. But in generally you should not do this, so please move | |
1882 | * along, nothing to see here.. | |
1883 | */ | |
1884 | struct file *filp_open(const char *filename, int flags, int mode) | |
1885 | { | |
6e8341a1 | 1886 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1887 | } |
1888 | EXPORT_SYMBOL(filp_open); | |
1889 | ||
1da177e4 LT |
1890 | /** |
1891 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1892 | * @nd: nameidata info | |
1893 | * @is_dir: directory flag | |
1894 | * | |
1895 | * Simple function to lookup and return a dentry and create it | |
1896 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1897 | * |
4ac91378 | 1898 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1899 | */ |
1900 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1901 | { | |
c663e5d8 | 1902 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1903 | |
4ac91378 | 1904 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1905 | /* |
1906 | * Yucky last component or no last component at all? | |
1907 | * (foo/., foo/.., /////) | |
1908 | */ | |
1da177e4 LT |
1909 | if (nd->last_type != LAST_NORM) |
1910 | goto fail; | |
1911 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1912 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1913 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1914 | |
1915 | /* | |
1916 | * Do the final lookup. | |
1917 | */ | |
49705b77 | 1918 | dentry = lookup_hash(nd); |
1da177e4 LT |
1919 | if (IS_ERR(dentry)) |
1920 | goto fail; | |
c663e5d8 | 1921 | |
e9baf6e5 AV |
1922 | if (dentry->d_inode) |
1923 | goto eexist; | |
c663e5d8 CH |
1924 | /* |
1925 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1926 | * From the vfs_mknod() POV we just have a negative dentry - | |
1927 | * all is fine. Let's be bastards - you had / on the end, you've | |
1928 | * been asking for (non-existent) directory. -ENOENT for you. | |
1929 | */ | |
e9baf6e5 AV |
1930 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1931 | dput(dentry); | |
1932 | dentry = ERR_PTR(-ENOENT); | |
1933 | } | |
1da177e4 | 1934 | return dentry; |
e9baf6e5 | 1935 | eexist: |
1da177e4 | 1936 | dput(dentry); |
e9baf6e5 | 1937 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1938 | fail: |
1939 | return dentry; | |
1940 | } | |
f81a0bff | 1941 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1942 | |
1943 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1944 | { | |
a95164d9 | 1945 | int error = may_create(dir, dentry); |
1da177e4 LT |
1946 | |
1947 | if (error) | |
1948 | return error; | |
1949 | ||
1950 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1951 | return -EPERM; | |
1952 | ||
acfa4380 | 1953 | if (!dir->i_op->mknod) |
1da177e4 LT |
1954 | return -EPERM; |
1955 | ||
08ce5f16 SH |
1956 | error = devcgroup_inode_mknod(mode, dev); |
1957 | if (error) | |
1958 | return error; | |
1959 | ||
1da177e4 LT |
1960 | error = security_inode_mknod(dir, dentry, mode, dev); |
1961 | if (error) | |
1962 | return error; | |
1963 | ||
9e3509e2 | 1964 | vfs_dq_init(dir); |
1da177e4 | 1965 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1966 | if (!error) |
f38aa942 | 1967 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1968 | return error; |
1969 | } | |
1970 | ||
463c3197 DH |
1971 | static int may_mknod(mode_t mode) |
1972 | { | |
1973 | switch (mode & S_IFMT) { | |
1974 | case S_IFREG: | |
1975 | case S_IFCHR: | |
1976 | case S_IFBLK: | |
1977 | case S_IFIFO: | |
1978 | case S_IFSOCK: | |
1979 | case 0: /* zero mode translates to S_IFREG */ | |
1980 | return 0; | |
1981 | case S_IFDIR: | |
1982 | return -EPERM; | |
1983 | default: | |
1984 | return -EINVAL; | |
1985 | } | |
1986 | } | |
1987 | ||
2e4d0924 HC |
1988 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
1989 | unsigned, dev) | |
1da177e4 | 1990 | { |
2ad94ae6 AV |
1991 | int error; |
1992 | char *tmp; | |
1993 | struct dentry *dentry; | |
1da177e4 LT |
1994 | struct nameidata nd; |
1995 | ||
1996 | if (S_ISDIR(mode)) | |
1997 | return -EPERM; | |
1da177e4 | 1998 | |
2ad94ae6 | 1999 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2000 | if (error) |
2ad94ae6 AV |
2001 | return error; |
2002 | ||
1da177e4 | 2003 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2004 | if (IS_ERR(dentry)) { |
2005 | error = PTR_ERR(dentry); | |
2006 | goto out_unlock; | |
2007 | } | |
4ac91378 | 2008 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2009 | mode &= ~current_umask(); |
463c3197 DH |
2010 | error = may_mknod(mode); |
2011 | if (error) | |
2012 | goto out_dput; | |
2013 | error = mnt_want_write(nd.path.mnt); | |
2014 | if (error) | |
2015 | goto out_dput; | |
be6d3e56 KT |
2016 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2017 | if (error) | |
2018 | goto out_drop_write; | |
463c3197 | 2019 | switch (mode & S_IFMT) { |
1da177e4 | 2020 | case 0: case S_IFREG: |
4ac91378 | 2021 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2022 | break; |
2023 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2024 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2025 | new_decode_dev(dev)); |
2026 | break; | |
2027 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2028 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2029 | break; |
1da177e4 | 2030 | } |
be6d3e56 | 2031 | out_drop_write: |
463c3197 DH |
2032 | mnt_drop_write(nd.path.mnt); |
2033 | out_dput: | |
2034 | dput(dentry); | |
2035 | out_unlock: | |
4ac91378 | 2036 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2037 | path_put(&nd.path); |
1da177e4 LT |
2038 | putname(tmp); |
2039 | ||
2040 | return error; | |
2041 | } | |
2042 | ||
3480b257 | 2043 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2044 | { |
2045 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2046 | } | |
2047 | ||
1da177e4 LT |
2048 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2049 | { | |
a95164d9 | 2050 | int error = may_create(dir, dentry); |
1da177e4 LT |
2051 | |
2052 | if (error) | |
2053 | return error; | |
2054 | ||
acfa4380 | 2055 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2056 | return -EPERM; |
2057 | ||
2058 | mode &= (S_IRWXUGO|S_ISVTX); | |
2059 | error = security_inode_mkdir(dir, dentry, mode); | |
2060 | if (error) | |
2061 | return error; | |
2062 | ||
9e3509e2 | 2063 | vfs_dq_init(dir); |
1da177e4 | 2064 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2065 | if (!error) |
f38aa942 | 2066 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2067 | return error; |
2068 | } | |
2069 | ||
2e4d0924 | 2070 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2071 | { |
2072 | int error = 0; | |
2073 | char * tmp; | |
6902d925 DH |
2074 | struct dentry *dentry; |
2075 | struct nameidata nd; | |
1da177e4 | 2076 | |
2ad94ae6 AV |
2077 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2078 | if (error) | |
6902d925 | 2079 | goto out_err; |
1da177e4 | 2080 | |
6902d925 DH |
2081 | dentry = lookup_create(&nd, 1); |
2082 | error = PTR_ERR(dentry); | |
2083 | if (IS_ERR(dentry)) | |
2084 | goto out_unlock; | |
1da177e4 | 2085 | |
4ac91378 | 2086 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2087 | mode &= ~current_umask(); |
463c3197 DH |
2088 | error = mnt_want_write(nd.path.mnt); |
2089 | if (error) | |
2090 | goto out_dput; | |
be6d3e56 KT |
2091 | error = security_path_mkdir(&nd.path, dentry, mode); |
2092 | if (error) | |
2093 | goto out_drop_write; | |
4ac91378 | 2094 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2095 | out_drop_write: |
463c3197 DH |
2096 | mnt_drop_write(nd.path.mnt); |
2097 | out_dput: | |
6902d925 DH |
2098 | dput(dentry); |
2099 | out_unlock: | |
4ac91378 | 2100 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2101 | path_put(&nd.path); |
6902d925 DH |
2102 | putname(tmp); |
2103 | out_err: | |
1da177e4 LT |
2104 | return error; |
2105 | } | |
2106 | ||
3cdad428 | 2107 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2108 | { |
2109 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2110 | } | |
2111 | ||
1da177e4 LT |
2112 | /* |
2113 | * We try to drop the dentry early: we should have | |
2114 | * a usage count of 2 if we're the only user of this | |
2115 | * dentry, and if that is true (possibly after pruning | |
2116 | * the dcache), then we drop the dentry now. | |
2117 | * | |
2118 | * A low-level filesystem can, if it choses, legally | |
2119 | * do a | |
2120 | * | |
2121 | * if (!d_unhashed(dentry)) | |
2122 | * return -EBUSY; | |
2123 | * | |
2124 | * if it cannot handle the case of removing a directory | |
2125 | * that is still in use by something else.. | |
2126 | */ | |
2127 | void dentry_unhash(struct dentry *dentry) | |
2128 | { | |
2129 | dget(dentry); | |
dc168427 | 2130 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2131 | spin_lock(&dcache_lock); |
2132 | spin_lock(&dentry->d_lock); | |
2133 | if (atomic_read(&dentry->d_count) == 2) | |
2134 | __d_drop(dentry); | |
2135 | spin_unlock(&dentry->d_lock); | |
2136 | spin_unlock(&dcache_lock); | |
2137 | } | |
2138 | ||
2139 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2140 | { | |
2141 | int error = may_delete(dir, dentry, 1); | |
2142 | ||
2143 | if (error) | |
2144 | return error; | |
2145 | ||
acfa4380 | 2146 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2147 | return -EPERM; |
2148 | ||
9e3509e2 | 2149 | vfs_dq_init(dir); |
1da177e4 | 2150 | |
1b1dcc1b | 2151 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2152 | dentry_unhash(dentry); |
2153 | if (d_mountpoint(dentry)) | |
2154 | error = -EBUSY; | |
2155 | else { | |
2156 | error = security_inode_rmdir(dir, dentry); | |
2157 | if (!error) { | |
2158 | error = dir->i_op->rmdir(dir, dentry); | |
2159 | if (!error) | |
2160 | dentry->d_inode->i_flags |= S_DEAD; | |
2161 | } | |
2162 | } | |
1b1dcc1b | 2163 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2164 | if (!error) { |
1da177e4 LT |
2165 | d_delete(dentry); |
2166 | } | |
2167 | dput(dentry); | |
2168 | ||
2169 | return error; | |
2170 | } | |
2171 | ||
5590ff0d | 2172 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2173 | { |
2174 | int error = 0; | |
2175 | char * name; | |
2176 | struct dentry *dentry; | |
2177 | struct nameidata nd; | |
2178 | ||
2ad94ae6 | 2179 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2180 | if (error) |
2ad94ae6 | 2181 | return error; |
1da177e4 LT |
2182 | |
2183 | switch(nd.last_type) { | |
0612d9fb OH |
2184 | case LAST_DOTDOT: |
2185 | error = -ENOTEMPTY; | |
2186 | goto exit1; | |
2187 | case LAST_DOT: | |
2188 | error = -EINVAL; | |
2189 | goto exit1; | |
2190 | case LAST_ROOT: | |
2191 | error = -EBUSY; | |
2192 | goto exit1; | |
1da177e4 | 2193 | } |
0612d9fb OH |
2194 | |
2195 | nd.flags &= ~LOOKUP_PARENT; | |
2196 | ||
4ac91378 | 2197 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2198 | dentry = lookup_hash(&nd); |
1da177e4 | 2199 | error = PTR_ERR(dentry); |
6902d925 DH |
2200 | if (IS_ERR(dentry)) |
2201 | goto exit2; | |
0622753b DH |
2202 | error = mnt_want_write(nd.path.mnt); |
2203 | if (error) | |
2204 | goto exit3; | |
be6d3e56 KT |
2205 | error = security_path_rmdir(&nd.path, dentry); |
2206 | if (error) | |
2207 | goto exit4; | |
4ac91378 | 2208 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2209 | exit4: |
0622753b DH |
2210 | mnt_drop_write(nd.path.mnt); |
2211 | exit3: | |
6902d925 DH |
2212 | dput(dentry); |
2213 | exit2: | |
4ac91378 | 2214 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2215 | exit1: |
1d957f9b | 2216 | path_put(&nd.path); |
1da177e4 LT |
2217 | putname(name); |
2218 | return error; | |
2219 | } | |
2220 | ||
3cdad428 | 2221 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2222 | { |
2223 | return do_rmdir(AT_FDCWD, pathname); | |
2224 | } | |
2225 | ||
1da177e4 LT |
2226 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2227 | { | |
2228 | int error = may_delete(dir, dentry, 0); | |
2229 | ||
2230 | if (error) | |
2231 | return error; | |
2232 | ||
acfa4380 | 2233 | if (!dir->i_op->unlink) |
1da177e4 LT |
2234 | return -EPERM; |
2235 | ||
9e3509e2 | 2236 | vfs_dq_init(dir); |
1da177e4 | 2237 | |
1b1dcc1b | 2238 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2239 | if (d_mountpoint(dentry)) |
2240 | error = -EBUSY; | |
2241 | else { | |
2242 | error = security_inode_unlink(dir, dentry); | |
2243 | if (!error) | |
2244 | error = dir->i_op->unlink(dir, dentry); | |
2245 | } | |
1b1dcc1b | 2246 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2247 | |
2248 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2249 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2250 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2251 | d_delete(dentry); |
1da177e4 | 2252 | } |
0eeca283 | 2253 | |
1da177e4 LT |
2254 | return error; |
2255 | } | |
2256 | ||
2257 | /* | |
2258 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2259 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2260 | * writeout happening, and we don't want to prevent access to the directory |
2261 | * while waiting on the I/O. | |
2262 | */ | |
5590ff0d | 2263 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2264 | { |
2ad94ae6 AV |
2265 | int error; |
2266 | char *name; | |
1da177e4 LT |
2267 | struct dentry *dentry; |
2268 | struct nameidata nd; | |
2269 | struct inode *inode = NULL; | |
2270 | ||
2ad94ae6 | 2271 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2272 | if (error) |
2ad94ae6 AV |
2273 | return error; |
2274 | ||
1da177e4 LT |
2275 | error = -EISDIR; |
2276 | if (nd.last_type != LAST_NORM) | |
2277 | goto exit1; | |
0612d9fb OH |
2278 | |
2279 | nd.flags &= ~LOOKUP_PARENT; | |
2280 | ||
4ac91378 | 2281 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2282 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2283 | error = PTR_ERR(dentry); |
2284 | if (!IS_ERR(dentry)) { | |
2285 | /* Why not before? Because we want correct error value */ | |
2286 | if (nd.last.name[nd.last.len]) | |
2287 | goto slashes; | |
2288 | inode = dentry->d_inode; | |
2289 | if (inode) | |
2290 | atomic_inc(&inode->i_count); | |
0622753b DH |
2291 | error = mnt_want_write(nd.path.mnt); |
2292 | if (error) | |
2293 | goto exit2; | |
be6d3e56 KT |
2294 | error = security_path_unlink(&nd.path, dentry); |
2295 | if (error) | |
2296 | goto exit3; | |
4ac91378 | 2297 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2298 | exit3: |
0622753b | 2299 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2300 | exit2: |
2301 | dput(dentry); | |
2302 | } | |
4ac91378 | 2303 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2304 | if (inode) |
2305 | iput(inode); /* truncate the inode here */ | |
2306 | exit1: | |
1d957f9b | 2307 | path_put(&nd.path); |
1da177e4 LT |
2308 | putname(name); |
2309 | return error; | |
2310 | ||
2311 | slashes: | |
2312 | error = !dentry->d_inode ? -ENOENT : | |
2313 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2314 | goto exit2; | |
2315 | } | |
2316 | ||
2e4d0924 | 2317 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2318 | { |
2319 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2320 | return -EINVAL; | |
2321 | ||
2322 | if (flag & AT_REMOVEDIR) | |
2323 | return do_rmdir(dfd, pathname); | |
2324 | ||
2325 | return do_unlinkat(dfd, pathname); | |
2326 | } | |
2327 | ||
3480b257 | 2328 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2329 | { |
2330 | return do_unlinkat(AT_FDCWD, pathname); | |
2331 | } | |
2332 | ||
db2e747b | 2333 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2334 | { |
a95164d9 | 2335 | int error = may_create(dir, dentry); |
1da177e4 LT |
2336 | |
2337 | if (error) | |
2338 | return error; | |
2339 | ||
acfa4380 | 2340 | if (!dir->i_op->symlink) |
1da177e4 LT |
2341 | return -EPERM; |
2342 | ||
2343 | error = security_inode_symlink(dir, dentry, oldname); | |
2344 | if (error) | |
2345 | return error; | |
2346 | ||
9e3509e2 | 2347 | vfs_dq_init(dir); |
1da177e4 | 2348 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2349 | if (!error) |
f38aa942 | 2350 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2351 | return error; |
2352 | } | |
2353 | ||
2e4d0924 HC |
2354 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2355 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2356 | { |
2ad94ae6 AV |
2357 | int error; |
2358 | char *from; | |
2359 | char *to; | |
6902d925 DH |
2360 | struct dentry *dentry; |
2361 | struct nameidata nd; | |
1da177e4 LT |
2362 | |
2363 | from = getname(oldname); | |
2ad94ae6 | 2364 | if (IS_ERR(from)) |
1da177e4 | 2365 | return PTR_ERR(from); |
1da177e4 | 2366 | |
2ad94ae6 | 2367 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2368 | if (error) |
2ad94ae6 AV |
2369 | goto out_putname; |
2370 | ||
6902d925 DH |
2371 | dentry = lookup_create(&nd, 0); |
2372 | error = PTR_ERR(dentry); | |
2373 | if (IS_ERR(dentry)) | |
2374 | goto out_unlock; | |
2375 | ||
75c3f29d DH |
2376 | error = mnt_want_write(nd.path.mnt); |
2377 | if (error) | |
2378 | goto out_dput; | |
be6d3e56 KT |
2379 | error = security_path_symlink(&nd.path, dentry, from); |
2380 | if (error) | |
2381 | goto out_drop_write; | |
db2e747b | 2382 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2383 | out_drop_write: |
75c3f29d DH |
2384 | mnt_drop_write(nd.path.mnt); |
2385 | out_dput: | |
6902d925 DH |
2386 | dput(dentry); |
2387 | out_unlock: | |
4ac91378 | 2388 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2389 | path_put(&nd.path); |
6902d925 DH |
2390 | putname(to); |
2391 | out_putname: | |
1da177e4 LT |
2392 | putname(from); |
2393 | return error; | |
2394 | } | |
2395 | ||
3480b257 | 2396 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2397 | { |
2398 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2399 | } | |
2400 | ||
1da177e4 LT |
2401 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2402 | { | |
2403 | struct inode *inode = old_dentry->d_inode; | |
2404 | int error; | |
2405 | ||
2406 | if (!inode) | |
2407 | return -ENOENT; | |
2408 | ||
a95164d9 | 2409 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2410 | if (error) |
2411 | return error; | |
2412 | ||
2413 | if (dir->i_sb != inode->i_sb) | |
2414 | return -EXDEV; | |
2415 | ||
2416 | /* | |
2417 | * A link to an append-only or immutable file cannot be created. | |
2418 | */ | |
2419 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2420 | return -EPERM; | |
acfa4380 | 2421 | if (!dir->i_op->link) |
1da177e4 | 2422 | return -EPERM; |
7e79eedb | 2423 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2424 | return -EPERM; |
2425 | ||
2426 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2427 | if (error) | |
2428 | return error; | |
2429 | ||
7e79eedb | 2430 | mutex_lock(&inode->i_mutex); |
9e3509e2 | 2431 | vfs_dq_init(dir); |
1da177e4 | 2432 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2433 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2434 | if (!error) |
7e79eedb | 2435 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2436 | return error; |
2437 | } | |
2438 | ||
2439 | /* | |
2440 | * Hardlinks are often used in delicate situations. We avoid | |
2441 | * security-related surprises by not following symlinks on the | |
2442 | * newname. --KAB | |
2443 | * | |
2444 | * We don't follow them on the oldname either to be compatible | |
2445 | * with linux 2.0, and to avoid hard-linking to directories | |
2446 | * and other special files. --ADM | |
2447 | */ | |
2e4d0924 HC |
2448 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2449 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2450 | { |
2451 | struct dentry *new_dentry; | |
2d8f3038 AV |
2452 | struct nameidata nd; |
2453 | struct path old_path; | |
1da177e4 | 2454 | int error; |
2ad94ae6 | 2455 | char *to; |
1da177e4 | 2456 | |
45c9b11a | 2457 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2458 | return -EINVAL; |
2459 | ||
2d8f3038 AV |
2460 | error = user_path_at(olddfd, oldname, |
2461 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2462 | &old_path); | |
1da177e4 | 2463 | if (error) |
2ad94ae6 AV |
2464 | return error; |
2465 | ||
2466 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2467 | if (error) |
2468 | goto out; | |
2469 | error = -EXDEV; | |
2d8f3038 | 2470 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2471 | goto out_release; |
2472 | new_dentry = lookup_create(&nd, 0); | |
2473 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2474 | if (IS_ERR(new_dentry)) |
2475 | goto out_unlock; | |
75c3f29d DH |
2476 | error = mnt_want_write(nd.path.mnt); |
2477 | if (error) | |
2478 | goto out_dput; | |
be6d3e56 KT |
2479 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2480 | if (error) | |
2481 | goto out_drop_write; | |
2d8f3038 | 2482 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2483 | out_drop_write: |
75c3f29d DH |
2484 | mnt_drop_write(nd.path.mnt); |
2485 | out_dput: | |
6902d925 DH |
2486 | dput(new_dentry); |
2487 | out_unlock: | |
4ac91378 | 2488 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2489 | out_release: |
1d957f9b | 2490 | path_put(&nd.path); |
2ad94ae6 | 2491 | putname(to); |
1da177e4 | 2492 | out: |
2d8f3038 | 2493 | path_put(&old_path); |
1da177e4 LT |
2494 | |
2495 | return error; | |
2496 | } | |
2497 | ||
3480b257 | 2498 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2499 | { |
c04030e1 | 2500 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2501 | } |
2502 | ||
1da177e4 LT |
2503 | /* |
2504 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2505 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2506 | * Problems: | |
2507 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2508 | * b) race potential - two innocent renames can create a loop together. | |
2509 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2510 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2511 | * story. |
2512 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2513 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2514 | * whether the target exists). Solution: try to be smart with locking |
2515 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2516 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2517 | * move will be locked. Thus we can rank directories by the tree |
2518 | * (ancestors first) and rank all non-directories after them. | |
2519 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2520 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2521 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2522 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2523 | * we'd better make sure that there's no link(2) for them. | |
2524 | * d) some filesystems don't support opened-but-unlinked directories, | |
2525 | * either because of layout or because they are not ready to deal with | |
2526 | * all cases correctly. The latter will be fixed (taking this sort of | |
2527 | * stuff into VFS), but the former is not going away. Solution: the same | |
2528 | * trick as in rmdir(). | |
2529 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2530 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2531 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2532 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2533 | * locking]. |
2534 | */ | |
75c96f85 AB |
2535 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2536 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2537 | { |
2538 | int error = 0; | |
2539 | struct inode *target; | |
2540 | ||
2541 | /* | |
2542 | * If we are going to change the parent - check write permissions, | |
2543 | * we'll need to flip '..'. | |
2544 | */ | |
2545 | if (new_dir != old_dir) { | |
f419a2e3 | 2546 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2547 | if (error) |
2548 | return error; | |
2549 | } | |
2550 | ||
2551 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2552 | if (error) | |
2553 | return error; | |
2554 | ||
2555 | target = new_dentry->d_inode; | |
2556 | if (target) { | |
1b1dcc1b | 2557 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2558 | dentry_unhash(new_dentry); |
2559 | } | |
2560 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2561 | error = -EBUSY; | |
2562 | else | |
2563 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2564 | if (target) { | |
2565 | if (!error) | |
2566 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2567 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2568 | if (d_unhashed(new_dentry)) |
2569 | d_rehash(new_dentry); | |
2570 | dput(new_dentry); | |
2571 | } | |
e31e14ec | 2572 | if (!error) |
349457cc MF |
2573 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2574 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2575 | return error; |
2576 | } | |
2577 | ||
75c96f85 AB |
2578 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2579 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2580 | { |
2581 | struct inode *target; | |
2582 | int error; | |
2583 | ||
2584 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2585 | if (error) | |
2586 | return error; | |
2587 | ||
2588 | dget(new_dentry); | |
2589 | target = new_dentry->d_inode; | |
2590 | if (target) | |
1b1dcc1b | 2591 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2592 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2593 | error = -EBUSY; | |
2594 | else | |
2595 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2596 | if (!error) { | |
349457cc | 2597 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2598 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2599 | } |
2600 | if (target) | |
1b1dcc1b | 2601 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2602 | dput(new_dentry); |
2603 | return error; | |
2604 | } | |
2605 | ||
2606 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2607 | struct inode *new_dir, struct dentry *new_dentry) | |
2608 | { | |
2609 | int error; | |
2610 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2611 | const char *old_name; |
1da177e4 LT |
2612 | |
2613 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2614 | return 0; | |
2615 | ||
2616 | error = may_delete(old_dir, old_dentry, is_dir); | |
2617 | if (error) | |
2618 | return error; | |
2619 | ||
2620 | if (!new_dentry->d_inode) | |
a95164d9 | 2621 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2622 | else |
2623 | error = may_delete(new_dir, new_dentry, is_dir); | |
2624 | if (error) | |
2625 | return error; | |
2626 | ||
acfa4380 | 2627 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2628 | return -EPERM; |
2629 | ||
9e3509e2 JK |
2630 | vfs_dq_init(old_dir); |
2631 | vfs_dq_init(new_dir); | |
1da177e4 | 2632 | |
0eeca283 RL |
2633 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2634 | ||
1da177e4 LT |
2635 | if (is_dir) |
2636 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2637 | else | |
2638 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2639 | if (!error) { | |
0eeca283 | 2640 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2641 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2642 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2643 | } |
0eeca283 RL |
2644 | fsnotify_oldname_free(old_name); |
2645 | ||
1da177e4 LT |
2646 | return error; |
2647 | } | |
2648 | ||
2e4d0924 HC |
2649 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2650 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2651 | { |
2ad94ae6 AV |
2652 | struct dentry *old_dir, *new_dir; |
2653 | struct dentry *old_dentry, *new_dentry; | |
2654 | struct dentry *trap; | |
1da177e4 | 2655 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2656 | char *from; |
2657 | char *to; | |
2658 | int error; | |
1da177e4 | 2659 | |
2ad94ae6 | 2660 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2661 | if (error) |
2662 | goto exit; | |
2663 | ||
2ad94ae6 | 2664 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2665 | if (error) |
2666 | goto exit1; | |
2667 | ||
2668 | error = -EXDEV; | |
4ac91378 | 2669 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2670 | goto exit2; |
2671 | ||
4ac91378 | 2672 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2673 | error = -EBUSY; |
2674 | if (oldnd.last_type != LAST_NORM) | |
2675 | goto exit2; | |
2676 | ||
4ac91378 | 2677 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2678 | if (newnd.last_type != LAST_NORM) |
2679 | goto exit2; | |
2680 | ||
0612d9fb OH |
2681 | oldnd.flags &= ~LOOKUP_PARENT; |
2682 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2683 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2684 | |
1da177e4 LT |
2685 | trap = lock_rename(new_dir, old_dir); |
2686 | ||
49705b77 | 2687 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2688 | error = PTR_ERR(old_dentry); |
2689 | if (IS_ERR(old_dentry)) | |
2690 | goto exit3; | |
2691 | /* source must exist */ | |
2692 | error = -ENOENT; | |
2693 | if (!old_dentry->d_inode) | |
2694 | goto exit4; | |
2695 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2696 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2697 | error = -ENOTDIR; | |
2698 | if (oldnd.last.name[oldnd.last.len]) | |
2699 | goto exit4; | |
2700 | if (newnd.last.name[newnd.last.len]) | |
2701 | goto exit4; | |
2702 | } | |
2703 | /* source should not be ancestor of target */ | |
2704 | error = -EINVAL; | |
2705 | if (old_dentry == trap) | |
2706 | goto exit4; | |
49705b77 | 2707 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2708 | error = PTR_ERR(new_dentry); |
2709 | if (IS_ERR(new_dentry)) | |
2710 | goto exit4; | |
2711 | /* target should not be an ancestor of source */ | |
2712 | error = -ENOTEMPTY; | |
2713 | if (new_dentry == trap) | |
2714 | goto exit5; | |
2715 | ||
9079b1eb DH |
2716 | error = mnt_want_write(oldnd.path.mnt); |
2717 | if (error) | |
2718 | goto exit5; | |
be6d3e56 KT |
2719 | error = security_path_rename(&oldnd.path, old_dentry, |
2720 | &newnd.path, new_dentry); | |
2721 | if (error) | |
2722 | goto exit6; | |
1da177e4 LT |
2723 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2724 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2725 | exit6: |
9079b1eb | 2726 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2727 | exit5: |
2728 | dput(new_dentry); | |
2729 | exit4: | |
2730 | dput(old_dentry); | |
2731 | exit3: | |
2732 | unlock_rename(new_dir, old_dir); | |
2733 | exit2: | |
1d957f9b | 2734 | path_put(&newnd.path); |
2ad94ae6 | 2735 | putname(to); |
1da177e4 | 2736 | exit1: |
1d957f9b | 2737 | path_put(&oldnd.path); |
1da177e4 | 2738 | putname(from); |
2ad94ae6 | 2739 | exit: |
1da177e4 LT |
2740 | return error; |
2741 | } | |
2742 | ||
a26eab24 | 2743 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2744 | { |
2745 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2746 | } | |
2747 | ||
1da177e4 LT |
2748 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2749 | { | |
2750 | int len; | |
2751 | ||
2752 | len = PTR_ERR(link); | |
2753 | if (IS_ERR(link)) | |
2754 | goto out; | |
2755 | ||
2756 | len = strlen(link); | |
2757 | if (len > (unsigned) buflen) | |
2758 | len = buflen; | |
2759 | if (copy_to_user(buffer, link, len)) | |
2760 | len = -EFAULT; | |
2761 | out: | |
2762 | return len; | |
2763 | } | |
2764 | ||
2765 | /* | |
2766 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2767 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2768 | * using) it for any given inode is up to filesystem. | |
2769 | */ | |
2770 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2771 | { | |
2772 | struct nameidata nd; | |
cc314eef | 2773 | void *cookie; |
694a1764 | 2774 | int res; |
cc314eef | 2775 | |
1da177e4 | 2776 | nd.depth = 0; |
cc314eef | 2777 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2778 | if (IS_ERR(cookie)) |
2779 | return PTR_ERR(cookie); | |
2780 | ||
2781 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2782 | if (dentry->d_inode->i_op->put_link) | |
2783 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2784 | return res; | |
1da177e4 LT |
2785 | } |
2786 | ||
2787 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2788 | { | |
2789 | return __vfs_follow_link(nd, link); | |
2790 | } | |
2791 | ||
2792 | /* get the link contents into pagecache */ | |
2793 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2794 | { | |
ebd09abb DG |
2795 | char *kaddr; |
2796 | struct page *page; | |
1da177e4 | 2797 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2798 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2799 | if (IS_ERR(page)) |
6fe6900e | 2800 | return (char*)page; |
1da177e4 | 2801 | *ppage = page; |
ebd09abb DG |
2802 | kaddr = kmap(page); |
2803 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2804 | return kaddr; | |
1da177e4 LT |
2805 | } |
2806 | ||
2807 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2808 | { | |
2809 | struct page *page = NULL; | |
2810 | char *s = page_getlink(dentry, &page); | |
2811 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2812 | if (page) { | |
2813 | kunmap(page); | |
2814 | page_cache_release(page); | |
2815 | } | |
2816 | return res; | |
2817 | } | |
2818 | ||
cc314eef | 2819 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2820 | { |
cc314eef | 2821 | struct page *page = NULL; |
1da177e4 | 2822 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2823 | return page; |
1da177e4 LT |
2824 | } |
2825 | ||
cc314eef | 2826 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2827 | { |
cc314eef LT |
2828 | struct page *page = cookie; |
2829 | ||
2830 | if (page) { | |
1da177e4 LT |
2831 | kunmap(page); |
2832 | page_cache_release(page); | |
1da177e4 LT |
2833 | } |
2834 | } | |
2835 | ||
54566b2c NP |
2836 | /* |
2837 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2838 | */ | |
2839 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2840 | { |
2841 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2842 | struct page *page; |
afddba49 | 2843 | void *fsdata; |
beb497ab | 2844 | int err; |
1da177e4 | 2845 | char *kaddr; |
54566b2c NP |
2846 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2847 | if (nofs) | |
2848 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2849 | |
7e53cac4 | 2850 | retry: |
afddba49 | 2851 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2852 | flags, &page, &fsdata); |
1da177e4 | 2853 | if (err) |
afddba49 NP |
2854 | goto fail; |
2855 | ||
1da177e4 LT |
2856 | kaddr = kmap_atomic(page, KM_USER0); |
2857 | memcpy(kaddr, symname, len-1); | |
2858 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2859 | |
2860 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2861 | page, fsdata); | |
1da177e4 LT |
2862 | if (err < 0) |
2863 | goto fail; | |
afddba49 NP |
2864 | if (err < len-1) |
2865 | goto retry; | |
2866 | ||
1da177e4 LT |
2867 | mark_inode_dirty(inode); |
2868 | return 0; | |
1da177e4 LT |
2869 | fail: |
2870 | return err; | |
2871 | } | |
2872 | ||
0adb25d2 KK |
2873 | int page_symlink(struct inode *inode, const char *symname, int len) |
2874 | { | |
2875 | return __page_symlink(inode, symname, len, | |
54566b2c | 2876 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2877 | } |
2878 | ||
92e1d5be | 2879 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2880 | .readlink = generic_readlink, |
2881 | .follow_link = page_follow_link_light, | |
2882 | .put_link = page_put_link, | |
2883 | }; | |
2884 | ||
2d8f3038 | 2885 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2886 | EXPORT_SYMBOL(follow_down); |
2887 | EXPORT_SYMBOL(follow_up); | |
2888 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2889 | EXPORT_SYMBOL(getname); | |
2890 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2891 | EXPORT_SYMBOL(lookup_one_len); |
2892 | EXPORT_SYMBOL(page_follow_link_light); | |
2893 | EXPORT_SYMBOL(page_put_link); | |
2894 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2895 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2896 | EXPORT_SYMBOL(page_symlink); |
2897 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2898 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2899 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2900 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2901 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2902 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2903 | EXPORT_SYMBOL(unlock_rename); |
2904 | EXPORT_SYMBOL(vfs_create); | |
2905 | EXPORT_SYMBOL(vfs_follow_link); | |
2906 | EXPORT_SYMBOL(vfs_link); | |
2907 | EXPORT_SYMBOL(vfs_mkdir); | |
2908 | EXPORT_SYMBOL(vfs_mknod); | |
2909 | EXPORT_SYMBOL(generic_permission); | |
2910 | EXPORT_SYMBOL(vfs_readlink); | |
2911 | EXPORT_SYMBOL(vfs_rename); | |
2912 | EXPORT_SYMBOL(vfs_rmdir); | |
2913 | EXPORT_SYMBOL(vfs_symlink); | |
2914 | EXPORT_SYMBOL(vfs_unlink); | |
2915 | EXPORT_SYMBOL(dentry_unhash); | |
2916 | EXPORT_SYMBOL(generic_readlink); |