]>
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) |
b7a437b0 | 438 | return inode_permission(inode, MAY_EXEC); |
1da177e4 | 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 | ||
2a737871 AV |
555 | static __always_inline void set_root(struct nameidata *nd) |
556 | { | |
557 | if (!nd->root.mnt) { | |
558 | struct fs_struct *fs = current->fs; | |
559 | read_lock(&fs->lock); | |
560 | nd->root = fs->root; | |
561 | path_get(&nd->root); | |
562 | read_unlock(&fs->lock); | |
563 | } | |
564 | } | |
565 | ||
f1662356 | 566 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
567 | { |
568 | int res = 0; | |
569 | char *name; | |
570 | if (IS_ERR(link)) | |
571 | goto fail; | |
572 | ||
573 | if (*link == '/') { | |
2a737871 | 574 | set_root(nd); |
1d957f9b | 575 | path_put(&nd->path); |
2a737871 AV |
576 | nd->path = nd->root; |
577 | path_get(&nd->root); | |
1da177e4 | 578 | } |
b4091d5f | 579 | |
1da177e4 | 580 | res = link_path_walk(link, nd); |
1da177e4 LT |
581 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
582 | return res; | |
583 | /* | |
584 | * If it is an iterative symlinks resolution in open_namei() we | |
585 | * have to copy the last component. And all that crap because of | |
586 | * bloody create() on broken symlinks. Furrfu... | |
587 | */ | |
588 | name = __getname(); | |
589 | if (unlikely(!name)) { | |
1d957f9b | 590 | path_put(&nd->path); |
1da177e4 LT |
591 | return -ENOMEM; |
592 | } | |
593 | strcpy(name, nd->last.name); | |
594 | nd->last.name = name; | |
595 | return 0; | |
596 | fail: | |
1d957f9b | 597 | path_put(&nd->path); |
1da177e4 LT |
598 | return PTR_ERR(link); |
599 | } | |
600 | ||
1d957f9b | 601 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
602 | { |
603 | dput(path->dentry); | |
4ac91378 | 604 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
605 | mntput(path->mnt); |
606 | } | |
607 | ||
608 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
609 | { | |
4ac91378 JB |
610 | dput(nd->path.dentry); |
611 | if (nd->path.mnt != path->mnt) | |
612 | mntput(nd->path.mnt); | |
613 | nd->path.mnt = path->mnt; | |
614 | nd->path.dentry = path->dentry; | |
051d3812 IK |
615 | } |
616 | ||
f1662356 | 617 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
618 | { |
619 | int error; | |
cc314eef | 620 | void *cookie; |
cd4e91d3 | 621 | struct dentry *dentry = path->dentry; |
1da177e4 | 622 | |
d671a1cb | 623 | touch_atime(path->mnt, dentry); |
1da177e4 | 624 | nd_set_link(nd, NULL); |
cd4e91d3 | 625 | |
4ac91378 | 626 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
627 | path_to_nameidata(path, nd); |
628 | dget(dentry); | |
629 | } | |
630 | mntget(path->mnt); | |
cc314eef LT |
631 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
632 | error = PTR_ERR(cookie); | |
633 | if (!IS_ERR(cookie)) { | |
1da177e4 | 634 | char *s = nd_get_link(nd); |
cc314eef | 635 | error = 0; |
1da177e4 LT |
636 | if (s) |
637 | error = __vfs_follow_link(nd, s); | |
638 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 639 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 640 | } |
09da5916 | 641 | path_put(path); |
1da177e4 LT |
642 | |
643 | return error; | |
644 | } | |
645 | ||
646 | /* | |
647 | * This limits recursive symlink follows to 8, while | |
648 | * limiting consecutive symlinks to 40. | |
649 | * | |
650 | * Without that kind of total limit, nasty chains of consecutive | |
651 | * symlinks can cause almost arbitrarily long lookups. | |
652 | */ | |
90ebe565 | 653 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
654 | { |
655 | int err = -ELOOP; | |
656 | if (current->link_count >= MAX_NESTED_LINKS) | |
657 | goto loop; | |
658 | if (current->total_link_count >= 40) | |
659 | goto loop; | |
660 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
661 | cond_resched(); | |
90ebe565 | 662 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
663 | if (err) |
664 | goto loop; | |
665 | current->link_count++; | |
666 | current->total_link_count++; | |
667 | nd->depth++; | |
cd4e91d3 | 668 | err = __do_follow_link(path, nd); |
839d9f93 AV |
669 | current->link_count--; |
670 | nd->depth--; | |
1da177e4 LT |
671 | return err; |
672 | loop: | |
1d957f9b JB |
673 | path_put_conditional(path, nd); |
674 | path_put(&nd->path); | |
1da177e4 LT |
675 | return err; |
676 | } | |
677 | ||
bab77ebf | 678 | int follow_up(struct path *path) |
1da177e4 LT |
679 | { |
680 | struct vfsmount *parent; | |
681 | struct dentry *mountpoint; | |
682 | spin_lock(&vfsmount_lock); | |
bab77ebf AV |
683 | parent = path->mnt->mnt_parent; |
684 | if (parent == path->mnt) { | |
1da177e4 LT |
685 | spin_unlock(&vfsmount_lock); |
686 | return 0; | |
687 | } | |
688 | mntget(parent); | |
bab77ebf | 689 | mountpoint = dget(path->mnt->mnt_mountpoint); |
1da177e4 | 690 | spin_unlock(&vfsmount_lock); |
bab77ebf AV |
691 | dput(path->dentry); |
692 | path->dentry = mountpoint; | |
693 | mntput(path->mnt); | |
694 | path->mnt = parent; | |
1da177e4 LT |
695 | return 1; |
696 | } | |
697 | ||
698 | /* no need for dcache_lock, as serialization is taken care in | |
699 | * namespace.c | |
700 | */ | |
463ffb2e AV |
701 | static int __follow_mount(struct path *path) |
702 | { | |
703 | int res = 0; | |
704 | while (d_mountpoint(path->dentry)) { | |
1c755af4 | 705 | struct vfsmount *mounted = lookup_mnt(path); |
463ffb2e AV |
706 | if (!mounted) |
707 | break; | |
708 | dput(path->dentry); | |
709 | if (res) | |
710 | mntput(path->mnt); | |
711 | path->mnt = mounted; | |
712 | path->dentry = dget(mounted->mnt_root); | |
713 | res = 1; | |
714 | } | |
715 | return res; | |
716 | } | |
717 | ||
79ed0226 | 718 | static void follow_mount(struct path *path) |
1da177e4 | 719 | { |
79ed0226 | 720 | while (d_mountpoint(path->dentry)) { |
1c755af4 | 721 | struct vfsmount *mounted = lookup_mnt(path); |
1da177e4 LT |
722 | if (!mounted) |
723 | break; | |
79ed0226 AV |
724 | dput(path->dentry); |
725 | mntput(path->mnt); | |
726 | path->mnt = mounted; | |
727 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 | 728 | } |
1da177e4 LT |
729 | } |
730 | ||
731 | /* no need for dcache_lock, as serialization is taken care in | |
732 | * namespace.c | |
733 | */ | |
9393bd07 | 734 | int follow_down(struct path *path) |
1da177e4 LT |
735 | { |
736 | struct vfsmount *mounted; | |
737 | ||
1c755af4 | 738 | mounted = lookup_mnt(path); |
1da177e4 | 739 | if (mounted) { |
9393bd07 AV |
740 | dput(path->dentry); |
741 | mntput(path->mnt); | |
742 | path->mnt = mounted; | |
743 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
744 | return 1; |
745 | } | |
746 | return 0; | |
747 | } | |
748 | ||
f1662356 | 749 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 750 | { |
2a737871 | 751 | set_root(nd); |
e518ddb7 | 752 | |
1da177e4 LT |
753 | while(1) { |
754 | struct vfsmount *parent; | |
4ac91378 | 755 | struct dentry *old = nd->path.dentry; |
1da177e4 | 756 | |
2a737871 AV |
757 | if (nd->path.dentry == nd->root.dentry && |
758 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
759 | break; |
760 | } | |
1da177e4 | 761 | spin_lock(&dcache_lock); |
4ac91378 JB |
762 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
763 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
764 | spin_unlock(&dcache_lock); |
765 | dput(old); | |
766 | break; | |
767 | } | |
768 | spin_unlock(&dcache_lock); | |
769 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
770 | parent = nd->path.mnt->mnt_parent; |
771 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
772 | spin_unlock(&vfsmount_lock); |
773 | break; | |
774 | } | |
775 | mntget(parent); | |
4ac91378 | 776 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
777 | spin_unlock(&vfsmount_lock); |
778 | dput(old); | |
4ac91378 JB |
779 | mntput(nd->path.mnt); |
780 | nd->path.mnt = parent; | |
1da177e4 | 781 | } |
79ed0226 | 782 | follow_mount(&nd->path); |
1da177e4 LT |
783 | } |
784 | ||
1da177e4 LT |
785 | /* |
786 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
787 | * small and for now I'd prefer to have fast path as straight as possible. | |
788 | * It _is_ time-critical. | |
789 | */ | |
790 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
791 | struct path *path) | |
792 | { | |
4ac91378 JB |
793 | struct vfsmount *mnt = nd->path.mnt; |
794 | struct dentry *dentry = __d_lookup(nd->path.dentry, name); | |
1da177e4 LT |
795 | |
796 | if (!dentry) | |
797 | goto need_lookup; | |
798 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
799 | goto need_revalidate; | |
800 | done: | |
801 | path->mnt = mnt; | |
802 | path->dentry = dentry; | |
634ee701 | 803 | __follow_mount(path); |
1da177e4 LT |
804 | return 0; |
805 | ||
806 | need_lookup: | |
4ac91378 | 807 | dentry = real_lookup(nd->path.dentry, name, nd); |
1da177e4 LT |
808 | if (IS_ERR(dentry)) |
809 | goto fail; | |
810 | goto done; | |
811 | ||
812 | need_revalidate: | |
bcdc5e01 IK |
813 | dentry = do_revalidate(dentry, nd); |
814 | if (!dentry) | |
815 | goto need_lookup; | |
816 | if (IS_ERR(dentry)) | |
817 | goto fail; | |
818 | goto done; | |
1da177e4 LT |
819 | |
820 | fail: | |
821 | return PTR_ERR(dentry); | |
822 | } | |
823 | ||
824 | /* | |
825 | * Name resolution. | |
ea3834d9 PM |
826 | * This is the basic name resolution function, turning a pathname into |
827 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 828 | * |
ea3834d9 PM |
829 | * Returns 0 and nd will have valid dentry and mnt on success. |
830 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 831 | */ |
fc9b52cd | 832 | static int __link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
833 | { |
834 | struct path next; | |
835 | struct inode *inode; | |
836 | int err; | |
837 | unsigned int lookup_flags = nd->flags; | |
838 | ||
839 | while (*name=='/') | |
840 | name++; | |
841 | if (!*name) | |
842 | goto return_reval; | |
843 | ||
4ac91378 | 844 | inode = nd->path.dentry->d_inode; |
1da177e4 | 845 | if (nd->depth) |
f55eab82 | 846 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
847 | |
848 | /* At this point we know we have a real path component. */ | |
849 | for(;;) { | |
850 | unsigned long hash; | |
851 | struct qstr this; | |
852 | unsigned int c; | |
853 | ||
cdce5d6b | 854 | nd->flags |= LOOKUP_CONTINUE; |
672b16b2 | 855 | err = exec_permission_lite(inode); |
1da177e4 LT |
856 | if (err) |
857 | break; | |
858 | ||
859 | this.name = name; | |
860 | c = *(const unsigned char *)name; | |
861 | ||
862 | hash = init_name_hash(); | |
863 | do { | |
864 | name++; | |
865 | hash = partial_name_hash(c, hash); | |
866 | c = *(const unsigned char *)name; | |
867 | } while (c && (c != '/')); | |
868 | this.len = name - (const char *) this.name; | |
869 | this.hash = end_name_hash(hash); | |
870 | ||
871 | /* remove trailing slashes? */ | |
872 | if (!c) | |
873 | goto last_component; | |
874 | while (*++name == '/'); | |
875 | if (!*name) | |
876 | goto last_with_slashes; | |
877 | ||
878 | /* | |
879 | * "." and ".." are special - ".." especially so because it has | |
880 | * to be able to know about the current root directory and | |
881 | * parent relationships. | |
882 | */ | |
883 | if (this.name[0] == '.') switch (this.len) { | |
884 | default: | |
885 | break; | |
886 | case 2: | |
887 | if (this.name[1] != '.') | |
888 | break; | |
58c465eb | 889 | follow_dotdot(nd); |
4ac91378 | 890 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
891 | /* fallthrough */ |
892 | case 1: | |
893 | continue; | |
894 | } | |
895 | /* | |
896 | * See if the low-level filesystem might want | |
897 | * to use its own hash.. | |
898 | */ | |
4ac91378 JB |
899 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
900 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
901 | &this); | |
1da177e4 LT |
902 | if (err < 0) |
903 | break; | |
904 | } | |
1da177e4 LT |
905 | /* This does the actual lookups.. */ |
906 | err = do_lookup(nd, &this, &next); | |
907 | if (err) | |
908 | break; | |
1da177e4 LT |
909 | |
910 | err = -ENOENT; | |
911 | inode = next.dentry->d_inode; | |
912 | if (!inode) | |
913 | goto out_dput; | |
1da177e4 LT |
914 | |
915 | if (inode->i_op->follow_link) { | |
90ebe565 | 916 | err = do_follow_link(&next, nd); |
1da177e4 LT |
917 | if (err) |
918 | goto return_err; | |
919 | err = -ENOENT; | |
4ac91378 | 920 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
921 | if (!inode) |
922 | break; | |
09dd17d3 MS |
923 | } else |
924 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
925 | err = -ENOTDIR; |
926 | if (!inode->i_op->lookup) | |
927 | break; | |
928 | continue; | |
929 | /* here ends the main loop */ | |
930 | ||
931 | last_with_slashes: | |
932 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
933 | last_component: | |
f55eab82 TM |
934 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
935 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
936 | if (lookup_flags & LOOKUP_PARENT) |
937 | goto lookup_parent; | |
938 | if (this.name[0] == '.') switch (this.len) { | |
939 | default: | |
940 | break; | |
941 | case 2: | |
942 | if (this.name[1] != '.') | |
943 | break; | |
58c465eb | 944 | follow_dotdot(nd); |
4ac91378 | 945 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
946 | /* fallthrough */ |
947 | case 1: | |
948 | goto return_reval; | |
949 | } | |
4ac91378 JB |
950 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { |
951 | err = nd->path.dentry->d_op->d_hash(nd->path.dentry, | |
952 | &this); | |
1da177e4 LT |
953 | if (err < 0) |
954 | break; | |
955 | } | |
956 | err = do_lookup(nd, &this, &next); | |
957 | if (err) | |
958 | break; | |
1da177e4 LT |
959 | inode = next.dentry->d_inode; |
960 | if ((lookup_flags & LOOKUP_FOLLOW) | |
acfa4380 | 961 | && inode && inode->i_op->follow_link) { |
90ebe565 | 962 | err = do_follow_link(&next, nd); |
1da177e4 LT |
963 | if (err) |
964 | goto return_err; | |
4ac91378 | 965 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
966 | } else |
967 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
968 | err = -ENOENT; |
969 | if (!inode) | |
970 | break; | |
971 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
972 | err = -ENOTDIR; | |
acfa4380 | 973 | if (!inode->i_op->lookup) |
1da177e4 LT |
974 | break; |
975 | } | |
976 | goto return_base; | |
977 | lookup_parent: | |
978 | nd->last = this; | |
979 | nd->last_type = LAST_NORM; | |
980 | if (this.name[0] != '.') | |
981 | goto return_base; | |
982 | if (this.len == 1) | |
983 | nd->last_type = LAST_DOT; | |
984 | else if (this.len == 2 && this.name[1] == '.') | |
985 | nd->last_type = LAST_DOTDOT; | |
986 | else | |
987 | goto return_base; | |
988 | return_reval: | |
989 | /* | |
990 | * We bypassed the ordinary revalidation routines. | |
991 | * We may need to check the cached dentry for staleness. | |
992 | */ | |
4ac91378 JB |
993 | if (nd->path.dentry && nd->path.dentry->d_sb && |
994 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
995 | err = -ESTALE; |
996 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
997 | if (!nd->path.dentry->d_op->d_revalidate( |
998 | nd->path.dentry, nd)) | |
1da177e4 LT |
999 | break; |
1000 | } | |
1001 | return_base: | |
1002 | return 0; | |
1003 | out_dput: | |
1d957f9b | 1004 | path_put_conditional(&next, nd); |
1da177e4 LT |
1005 | break; |
1006 | } | |
1d957f9b | 1007 | path_put(&nd->path); |
1da177e4 LT |
1008 | return_err: |
1009 | return err; | |
1010 | } | |
1011 | ||
fc9b52cd | 1012 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1013 | { |
1014 | current->total_link_count = 0; | |
1015 | return link_path_walk(name, nd); | |
1016 | } | |
1017 | ||
9b4a9b14 | 1018 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1019 | { |
ea3834d9 | 1020 | int retval = 0; |
170aa3d0 UD |
1021 | int fput_needed; |
1022 | struct file *file; | |
1da177e4 LT |
1023 | |
1024 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1025 | nd->flags = flags; | |
1026 | nd->depth = 0; | |
2a737871 | 1027 | nd->root.mnt = NULL; |
1da177e4 | 1028 | |
1da177e4 | 1029 | if (*name=='/') { |
2a737871 AV |
1030 | set_root(nd); |
1031 | nd->path = nd->root; | |
1032 | path_get(&nd->root); | |
5590ff0d | 1033 | } else if (dfd == AT_FDCWD) { |
2a737871 | 1034 | struct fs_struct *fs = current->fs; |
e518ddb7 | 1035 | read_lock(&fs->lock); |
6ac08c39 JB |
1036 | nd->path = fs->pwd; |
1037 | path_get(&fs->pwd); | |
e518ddb7 | 1038 | read_unlock(&fs->lock); |
5590ff0d | 1039 | } else { |
5590ff0d UD |
1040 | struct dentry *dentry; |
1041 | ||
1042 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1043 | retval = -EBADF; |
1044 | if (!file) | |
6d09bb62 | 1045 | goto out_fail; |
5590ff0d | 1046 | |
0f7fc9e4 | 1047 | dentry = file->f_path.dentry; |
5590ff0d | 1048 | |
170aa3d0 UD |
1049 | retval = -ENOTDIR; |
1050 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1051 | goto fput_fail; |
5590ff0d UD |
1052 | |
1053 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1054 | if (retval) |
6d09bb62 | 1055 | goto fput_fail; |
5590ff0d | 1056 | |
5dd784d0 JB |
1057 | nd->path = file->f_path; |
1058 | path_get(&file->f_path); | |
5590ff0d UD |
1059 | |
1060 | fput_light(file, fput_needed); | |
1da177e4 | 1061 | } |
9b4a9b14 | 1062 | return 0; |
2dfdd266 | 1063 | |
9b4a9b14 AV |
1064 | fput_fail: |
1065 | fput_light(file, fput_needed); | |
1066 | out_fail: | |
1067 | return retval; | |
1068 | } | |
1069 | ||
1070 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1071 | static int do_path_lookup(int dfd, const char *name, | |
1072 | unsigned int flags, struct nameidata *nd) | |
1073 | { | |
1074 | int retval = path_init(dfd, name, flags, nd); | |
1075 | if (!retval) | |
1076 | retval = path_walk(name, nd); | |
4ac91378 JB |
1077 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1078 | nd->path.dentry->d_inode)) | |
1079 | audit_inode(name, nd->path.dentry); | |
2a737871 AV |
1080 | if (nd->root.mnt) { |
1081 | path_put(&nd->root); | |
1082 | nd->root.mnt = NULL; | |
1083 | } | |
170aa3d0 | 1084 | return retval; |
1da177e4 LT |
1085 | } |
1086 | ||
fc9b52cd | 1087 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1088 | struct nameidata *nd) |
1089 | { | |
1090 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1091 | } | |
1092 | ||
d1811465 AV |
1093 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1094 | { | |
1095 | struct nameidata nd; | |
1096 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1097 | if (!res) | |
1098 | *path = nd.path; | |
1099 | return res; | |
1100 | } | |
1101 | ||
16f18200 JJS |
1102 | /** |
1103 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1104 | * @dentry: pointer to dentry of the base directory | |
1105 | * @mnt: pointer to vfs mount of the base directory | |
1106 | * @name: pointer to file name | |
1107 | * @flags: lookup flags | |
1108 | * @nd: pointer to nameidata | |
1109 | */ | |
1110 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1111 | const char *name, unsigned int flags, | |
1112 | struct nameidata *nd) | |
1113 | { | |
1114 | int retval; | |
1115 | ||
1116 | /* same as do_path_lookup */ | |
1117 | nd->last_type = LAST_ROOT; | |
1118 | nd->flags = flags; | |
1119 | nd->depth = 0; | |
1120 | ||
c8e7f449 JB |
1121 | nd->path.dentry = dentry; |
1122 | nd->path.mnt = mnt; | |
1123 | path_get(&nd->path); | |
5b857119 AV |
1124 | nd->root = nd->path; |
1125 | path_get(&nd->root); | |
16f18200 JJS |
1126 | |
1127 | retval = path_walk(name, nd); | |
4ac91378 JB |
1128 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1129 | nd->path.dentry->d_inode)) | |
1130 | audit_inode(name, nd->path.dentry); | |
16f18200 | 1131 | |
5b857119 AV |
1132 | path_put(&nd->root); |
1133 | nd->root.mnt = NULL; | |
16f18200 | 1134 | |
2a737871 | 1135 | return retval; |
16f18200 JJS |
1136 | } |
1137 | ||
8737f3a1 AV |
1138 | /** |
1139 | * path_lookup_open - lookup a file path with open intent | |
1140 | * @dfd: the directory to use as base, or AT_FDCWD | |
1141 | * @name: pointer to file name | |
1142 | * @lookup_flags: lookup intent flags | |
1143 | * @nd: pointer to nameidata | |
1144 | * @open_flags: open intent flags | |
1145 | */ | |
6e8341a1 AV |
1146 | static int path_lookup_open(int dfd, const char *name, |
1147 | unsigned int lookup_flags, struct nameidata *nd, int open_flags) | |
834f2a4a TM |
1148 | { |
1149 | struct file *filp = get_empty_filp(); | |
1150 | int err; | |
1151 | ||
1152 | if (filp == NULL) | |
1153 | return -ENFILE; | |
1154 | nd->intent.open.file = filp; | |
1155 | nd->intent.open.flags = open_flags; | |
8737f3a1 | 1156 | nd->intent.open.create_mode = 0; |
5590ff0d | 1157 | err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd); |
834f2a4a TM |
1158 | if (IS_ERR(nd->intent.open.file)) { |
1159 | if (err == 0) { | |
1160 | err = PTR_ERR(nd->intent.open.file); | |
1d957f9b | 1161 | path_put(&nd->path); |
834f2a4a TM |
1162 | } |
1163 | } else if (err != 0) | |
1164 | release_open_intent(nd); | |
1165 | return err; | |
1166 | } | |
1167 | ||
eead1911 CH |
1168 | static struct dentry *__lookup_hash(struct qstr *name, |
1169 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1170 | { |
057f6c01 | 1171 | struct dentry *dentry; |
1da177e4 LT |
1172 | struct inode *inode; |
1173 | int err; | |
1174 | ||
1175 | inode = base->d_inode; | |
1da177e4 LT |
1176 | |
1177 | /* | |
1178 | * See if the low-level filesystem might want | |
1179 | * to use its own hash.. | |
1180 | */ | |
1181 | if (base->d_op && base->d_op->d_hash) { | |
1182 | err = base->d_op->d_hash(base, name); | |
1183 | dentry = ERR_PTR(err); | |
1184 | if (err < 0) | |
1185 | goto out; | |
1186 | } | |
1187 | ||
1188 | dentry = cached_lookup(base, name, nd); | |
1189 | if (!dentry) { | |
d70b67c8 MS |
1190 | struct dentry *new; |
1191 | ||
1192 | /* Don't create child dentry for a dead directory. */ | |
1193 | dentry = ERR_PTR(-ENOENT); | |
1194 | if (IS_DEADDIR(inode)) | |
1195 | goto out; | |
1196 | ||
1197 | new = d_alloc(base, name); | |
1da177e4 LT |
1198 | dentry = ERR_PTR(-ENOMEM); |
1199 | if (!new) | |
1200 | goto out; | |
1201 | dentry = inode->i_op->lookup(inode, new, nd); | |
1202 | if (!dentry) | |
1203 | dentry = new; | |
1204 | else | |
1205 | dput(new); | |
1206 | } | |
1207 | out: | |
1208 | return dentry; | |
1209 | } | |
1210 | ||
057f6c01 JM |
1211 | /* |
1212 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1213 | * needs parent already locked. Doesn't follow mounts. | |
1214 | * SMP-safe. | |
1215 | */ | |
eead1911 | 1216 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1217 | { |
057f6c01 JM |
1218 | int err; |
1219 | ||
f419a2e3 | 1220 | err = inode_permission(nd->path.dentry->d_inode, MAY_EXEC); |
057f6c01 | 1221 | if (err) |
eead1911 | 1222 | return ERR_PTR(err); |
4ac91378 | 1223 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1224 | } |
1225 | ||
eead1911 CH |
1226 | static int __lookup_one_len(const char *name, struct qstr *this, |
1227 | struct dentry *base, int len) | |
1da177e4 LT |
1228 | { |
1229 | unsigned long hash; | |
1da177e4 LT |
1230 | unsigned int c; |
1231 | ||
057f6c01 JM |
1232 | this->name = name; |
1233 | this->len = len; | |
1da177e4 | 1234 | if (!len) |
057f6c01 | 1235 | return -EACCES; |
1da177e4 LT |
1236 | |
1237 | hash = init_name_hash(); | |
1238 | while (len--) { | |
1239 | c = *(const unsigned char *)name++; | |
1240 | if (c == '/' || c == '\0') | |
057f6c01 | 1241 | return -EACCES; |
1da177e4 LT |
1242 | hash = partial_name_hash(c, hash); |
1243 | } | |
057f6c01 JM |
1244 | this->hash = end_name_hash(hash); |
1245 | return 0; | |
1246 | } | |
1da177e4 | 1247 | |
eead1911 | 1248 | /** |
a6b91919 | 1249 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1250 | * @name: pathname component to lookup |
1251 | * @base: base directory to lookup from | |
1252 | * @len: maximum length @len should be interpreted to | |
1253 | * | |
a6b91919 RD |
1254 | * Note that this routine is purely a helper for filesystem usage and should |
1255 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1256 | * nameidata argument is passed to the filesystem methods and a filesystem |
1257 | * using this helper needs to be prepared for that. | |
1258 | */ | |
057f6c01 JM |
1259 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1260 | { | |
1261 | int err; | |
1262 | struct qstr this; | |
1263 | ||
2f9092e1 DW |
1264 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1265 | ||
057f6c01 | 1266 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1267 | if (err) |
1268 | return ERR_PTR(err); | |
1269 | ||
f419a2e3 | 1270 | err = inode_permission(base->d_inode, MAY_EXEC); |
057f6c01 JM |
1271 | if (err) |
1272 | return ERR_PTR(err); | |
49705b77 | 1273 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1274 | } |
1275 | ||
eead1911 CH |
1276 | /** |
1277 | * lookup_one_noperm - bad hack for sysfs | |
1278 | * @name: pathname component to lookup | |
1279 | * @base: base directory to lookup from | |
1280 | * | |
1281 | * This is a variant of lookup_one_len that doesn't perform any permission | |
1282 | * checks. It's a horrible hack to work around the braindead sysfs | |
1283 | * architecture and should not be used anywhere else. | |
1284 | * | |
1285 | * DON'T USE THIS FUNCTION EVER, thanks. | |
1286 | */ | |
1287 | struct dentry *lookup_one_noperm(const char *name, struct dentry *base) | |
057f6c01 JM |
1288 | { |
1289 | int err; | |
1290 | struct qstr this; | |
1291 | ||
eead1911 | 1292 | err = __lookup_one_len(name, &this, base, strlen(name)); |
057f6c01 JM |
1293 | if (err) |
1294 | return ERR_PTR(err); | |
eead1911 | 1295 | return __lookup_hash(&this, base, NULL); |
1da177e4 LT |
1296 | } |
1297 | ||
2d8f3038 AV |
1298 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1299 | struct path *path) | |
1da177e4 | 1300 | { |
2d8f3038 | 1301 | struct nameidata nd; |
1da177e4 LT |
1302 | char *tmp = getname(name); |
1303 | int err = PTR_ERR(tmp); | |
1da177e4 | 1304 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1305 | |
1306 | BUG_ON(flags & LOOKUP_PARENT); | |
1307 | ||
1308 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1309 | putname(tmp); |
2d8f3038 AV |
1310 | if (!err) |
1311 | *path = nd.path; | |
1da177e4 LT |
1312 | } |
1313 | return err; | |
1314 | } | |
1315 | ||
2ad94ae6 AV |
1316 | static int user_path_parent(int dfd, const char __user *path, |
1317 | struct nameidata *nd, char **name) | |
1318 | { | |
1319 | char *s = getname(path); | |
1320 | int error; | |
1321 | ||
1322 | if (IS_ERR(s)) | |
1323 | return PTR_ERR(s); | |
1324 | ||
1325 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1326 | if (error) | |
1327 | putname(s); | |
1328 | else | |
1329 | *name = s; | |
1330 | ||
1331 | return error; | |
1332 | } | |
1333 | ||
1da177e4 LT |
1334 | /* |
1335 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1336 | * minimal. | |
1337 | */ | |
1338 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1339 | { | |
da9592ed DH |
1340 | uid_t fsuid = current_fsuid(); |
1341 | ||
1da177e4 LT |
1342 | if (!(dir->i_mode & S_ISVTX)) |
1343 | return 0; | |
da9592ed | 1344 | if (inode->i_uid == fsuid) |
1da177e4 | 1345 | return 0; |
da9592ed | 1346 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1347 | return 0; |
1348 | return !capable(CAP_FOWNER); | |
1349 | } | |
1350 | ||
1351 | /* | |
1352 | * Check whether we can remove a link victim from directory dir, check | |
1353 | * whether the type of victim is right. | |
1354 | * 1. We can't do it if dir is read-only (done in permission()) | |
1355 | * 2. We should have write and exec permissions on dir | |
1356 | * 3. We can't remove anything from append-only dir | |
1357 | * 4. We can't do anything with immutable dir (done in permission()) | |
1358 | * 5. If the sticky bit on dir is set we should either | |
1359 | * a. be owner of dir, or | |
1360 | * b. be owner of victim, or | |
1361 | * c. have CAP_FOWNER capability | |
1362 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1363 | * links pointing to it. | |
1364 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1365 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1366 | * 9. We can't remove a root or mountpoint. | |
1367 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1368 | * nfs_async_unlink(). | |
1369 | */ | |
858119e1 | 1370 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1371 | { |
1372 | int error; | |
1373 | ||
1374 | if (!victim->d_inode) | |
1375 | return -ENOENT; | |
1376 | ||
1377 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1378 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1379 | |
f419a2e3 | 1380 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1381 | if (error) |
1382 | return error; | |
1383 | if (IS_APPEND(dir)) | |
1384 | return -EPERM; | |
1385 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1386 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1387 | return -EPERM; |
1388 | if (isdir) { | |
1389 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1390 | return -ENOTDIR; | |
1391 | if (IS_ROOT(victim)) | |
1392 | return -EBUSY; | |
1393 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1394 | return -EISDIR; | |
1395 | if (IS_DEADDIR(dir)) | |
1396 | return -ENOENT; | |
1397 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1398 | return -EBUSY; | |
1399 | return 0; | |
1400 | } | |
1401 | ||
1402 | /* Check whether we can create an object with dentry child in directory | |
1403 | * dir. | |
1404 | * 1. We can't do it if child already exists (open has special treatment for | |
1405 | * this case, but since we are inlined it's OK) | |
1406 | * 2. We can't do it if dir is read-only (done in permission()) | |
1407 | * 3. We should have write and exec permissions on dir | |
1408 | * 4. We can't do it if dir is immutable (done in permission()) | |
1409 | */ | |
a95164d9 | 1410 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1411 | { |
1412 | if (child->d_inode) | |
1413 | return -EEXIST; | |
1414 | if (IS_DEADDIR(dir)) | |
1415 | return -ENOENT; | |
f419a2e3 | 1416 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1417 | } |
1418 | ||
1419 | /* | |
1da177e4 LT |
1420 | * O_DIRECTORY translates into forcing a directory lookup. |
1421 | */ | |
1422 | static inline int lookup_flags(unsigned int f) | |
1423 | { | |
1424 | unsigned long retval = LOOKUP_FOLLOW; | |
1425 | ||
1426 | if (f & O_NOFOLLOW) | |
1427 | retval &= ~LOOKUP_FOLLOW; | |
1428 | ||
1da177e4 LT |
1429 | if (f & O_DIRECTORY) |
1430 | retval |= LOOKUP_DIRECTORY; | |
1431 | ||
1432 | return retval; | |
1433 | } | |
1434 | ||
1435 | /* | |
1436 | * p1 and p2 should be directories on the same fs. | |
1437 | */ | |
1438 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1439 | { | |
1440 | struct dentry *p; | |
1441 | ||
1442 | if (p1 == p2) { | |
f2eace23 | 1443 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1444 | return NULL; |
1445 | } | |
1446 | ||
a11f3a05 | 1447 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1448 | |
e2761a11 OH |
1449 | p = d_ancestor(p2, p1); |
1450 | if (p) { | |
1451 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1452 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1453 | return p; | |
1da177e4 LT |
1454 | } |
1455 | ||
e2761a11 OH |
1456 | p = d_ancestor(p1, p2); |
1457 | if (p) { | |
1458 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1459 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1460 | return p; | |
1da177e4 LT |
1461 | } |
1462 | ||
f2eace23 IM |
1463 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1464 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1465 | return NULL; |
1466 | } | |
1467 | ||
1468 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1469 | { | |
1b1dcc1b | 1470 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1471 | if (p1 != p2) { |
1b1dcc1b | 1472 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1473 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1474 | } |
1475 | } | |
1476 | ||
1477 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1478 | struct nameidata *nd) | |
1479 | { | |
a95164d9 | 1480 | int error = may_create(dir, dentry); |
1da177e4 LT |
1481 | |
1482 | if (error) | |
1483 | return error; | |
1484 | ||
acfa4380 | 1485 | if (!dir->i_op->create) |
1da177e4 LT |
1486 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1487 | mode &= S_IALLUGO; | |
1488 | mode |= S_IFREG; | |
1489 | error = security_inode_create(dir, dentry, mode); | |
1490 | if (error) | |
1491 | return error; | |
9e3509e2 | 1492 | vfs_dq_init(dir); |
1da177e4 | 1493 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1494 | if (!error) |
f38aa942 | 1495 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1496 | return error; |
1497 | } | |
1498 | ||
3fb64190 | 1499 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1500 | { |
3fb64190 | 1501 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1502 | struct inode *inode = dentry->d_inode; |
1503 | int error; | |
1504 | ||
1505 | if (!inode) | |
1506 | return -ENOENT; | |
1507 | ||
c8fe8f30 CH |
1508 | switch (inode->i_mode & S_IFMT) { |
1509 | case S_IFLNK: | |
1da177e4 | 1510 | return -ELOOP; |
c8fe8f30 CH |
1511 | case S_IFDIR: |
1512 | if (acc_mode & MAY_WRITE) | |
1513 | return -EISDIR; | |
1514 | break; | |
1515 | case S_IFBLK: | |
1516 | case S_IFCHR: | |
3fb64190 | 1517 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1518 | return -EACCES; |
c8fe8f30 CH |
1519 | /*FALLTHRU*/ |
1520 | case S_IFIFO: | |
1521 | case S_IFSOCK: | |
1da177e4 | 1522 | flag &= ~O_TRUNC; |
c8fe8f30 | 1523 | break; |
4a3fd211 | 1524 | } |
b41572e9 | 1525 | |
3fb64190 | 1526 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1527 | if (error) |
1528 | return error; | |
6146f0d5 | 1529 | |
cb5629b1 | 1530 | error = ima_path_check(path, |
b9fc745d MZ |
1531 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC), |
1532 | IMA_COUNT_UPDATE); | |
6146f0d5 MZ |
1533 | if (error) |
1534 | return error; | |
1da177e4 LT |
1535 | /* |
1536 | * An append-only file must be opened in append mode for writing. | |
1537 | */ | |
1538 | if (IS_APPEND(inode)) { | |
acd0c935 | 1539 | error = -EPERM; |
1da177e4 | 1540 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) |
acd0c935 | 1541 | goto err_out; |
1da177e4 | 1542 | if (flag & O_TRUNC) |
acd0c935 | 1543 | goto err_out; |
1da177e4 LT |
1544 | } |
1545 | ||
1546 | /* O_NOATIME can only be set by the owner or superuser */ | |
1547 | if (flag & O_NOATIME) | |
acd0c935 MZ |
1548 | if (!is_owner_or_cap(inode)) { |
1549 | error = -EPERM; | |
1550 | goto err_out; | |
1551 | } | |
1da177e4 LT |
1552 | |
1553 | /* | |
1554 | * Ensure there are no outstanding leases on the file. | |
1555 | */ | |
1556 | error = break_lease(inode, flag); | |
1557 | if (error) | |
acd0c935 | 1558 | goto err_out; |
1da177e4 LT |
1559 | |
1560 | if (flag & O_TRUNC) { | |
1561 | error = get_write_access(inode); | |
1562 | if (error) | |
acd0c935 | 1563 | goto err_out; |
1da177e4 LT |
1564 | |
1565 | /* | |
1566 | * Refuse to truncate files with mandatory locks held on them. | |
1567 | */ | |
1568 | error = locks_verify_locked(inode); | |
be6d3e56 | 1569 | if (!error) |
3fb64190 | 1570 | error = security_path_truncate(path, 0, |
be6d3e56 | 1571 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN); |
1da177e4 | 1572 | if (!error) { |
9e3509e2 | 1573 | vfs_dq_init(inode); |
d139d7ff MS |
1574 | |
1575 | error = do_truncate(dentry, 0, | |
1576 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1577 | NULL); | |
1da177e4 LT |
1578 | } |
1579 | put_write_access(inode); | |
1580 | if (error) | |
acd0c935 | 1581 | goto err_out; |
1da177e4 LT |
1582 | } else |
1583 | if (flag & FMODE_WRITE) | |
9e3509e2 | 1584 | vfs_dq_init(inode); |
1da177e4 LT |
1585 | |
1586 | return 0; | |
acd0c935 MZ |
1587 | err_out: |
1588 | ima_counts_put(path, acc_mode ? | |
1589 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC) : | |
1590 | ACC_MODE(flag) & (MAY_READ | MAY_WRITE)); | |
1591 | return error; | |
1da177e4 LT |
1592 | } |
1593 | ||
d57999e1 DH |
1594 | /* |
1595 | * Be careful about ever adding any more callers of this | |
1596 | * function. Its flags must be in the namei format, not | |
1597 | * what get passed to sys_open(). | |
1598 | */ | |
1599 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1600 | int flag, int mode) |
1601 | { | |
1602 | int error; | |
4ac91378 | 1603 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1604 | |
1605 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1606 | mode &= ~current_umask(); |
be6d3e56 KT |
1607 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1608 | if (error) | |
1609 | goto out_unlock; | |
aab520e2 | 1610 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1611 | out_unlock: |
aab520e2 | 1612 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1613 | dput(nd->path.dentry); |
1614 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1615 | if (error) |
1616 | return error; | |
1617 | /* Don't check for write permission, don't truncate */ | |
3fb64190 | 1618 | return may_open(&nd->path, 0, flag & ~O_TRUNC); |
aab520e2 DH |
1619 | } |
1620 | ||
d57999e1 DH |
1621 | /* |
1622 | * Note that while the flag value (low two bits) for sys_open means: | |
1623 | * 00 - read-only | |
1624 | * 01 - write-only | |
1625 | * 10 - read-write | |
1626 | * 11 - special | |
1627 | * it is changed into | |
1628 | * 00 - no permissions needed | |
1629 | * 01 - read-permission | |
1630 | * 10 - write-permission | |
1631 | * 11 - read-write | |
1632 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1633 | * This is more logical, and also allows the 00 "no perm needed" | |
1634 | * to be used for symlinks (where the permissions are checked | |
1635 | * later). | |
1636 | * | |
1637 | */ | |
1638 | static inline int open_to_namei_flags(int flag) | |
1639 | { | |
1640 | if ((flag+1) & O_ACCMODE) | |
1641 | flag++; | |
1642 | return flag; | |
1643 | } | |
1644 | ||
4a3fd211 DH |
1645 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1646 | { | |
1647 | /* | |
1648 | * We'll never write to the fs underlying | |
1649 | * a device file. | |
1650 | */ | |
1651 | if (special_file(inode->i_mode)) | |
1652 | return 0; | |
1653 | return (flag & O_TRUNC); | |
1654 | } | |
1655 | ||
1da177e4 | 1656 | /* |
4a3fd211 DH |
1657 | * Note that the low bits of the passed in "open_flag" |
1658 | * are not the same as in the local variable "flag". See | |
1659 | * open_to_namei_flags() for more details. | |
1da177e4 | 1660 | */ |
a70e65df | 1661 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1662 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1663 | { |
4a3fd211 | 1664 | struct file *filp; |
a70e65df | 1665 | struct nameidata nd; |
6e8341a1 | 1666 | int error; |
4e7506e4 | 1667 | struct path path; |
1da177e4 LT |
1668 | struct dentry *dir; |
1669 | int count = 0; | |
4a3fd211 | 1670 | int will_write; |
d57999e1 | 1671 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1672 | |
6e8341a1 AV |
1673 | if (!acc_mode) |
1674 | acc_mode = MAY_OPEN | ACC_MODE(flag); | |
1da177e4 | 1675 | |
834f2a4a TM |
1676 | /* O_TRUNC implies we need access checks for write permissions */ |
1677 | if (flag & O_TRUNC) | |
1678 | acc_mode |= MAY_WRITE; | |
1679 | ||
1da177e4 LT |
1680 | /* Allow the LSM permission hook to distinguish append |
1681 | access from general write access. */ | |
1682 | if (flag & O_APPEND) | |
1683 | acc_mode |= MAY_APPEND; | |
1684 | ||
1da177e4 LT |
1685 | /* |
1686 | * The simplest case - just a plain lookup. | |
1687 | */ | |
1688 | if (!(flag & O_CREAT)) { | |
5590ff0d | 1689 | error = path_lookup_open(dfd, pathname, lookup_flags(flag), |
a70e65df | 1690 | &nd, flag); |
1da177e4 | 1691 | if (error) |
a70e65df | 1692 | return ERR_PTR(error); |
1da177e4 LT |
1693 | goto ok; |
1694 | } | |
1695 | ||
1696 | /* | |
1697 | * Create - we need to know the parent. | |
1698 | */ | |
9b4a9b14 | 1699 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1700 | if (error) |
a70e65df | 1701 | return ERR_PTR(error); |
9b4a9b14 | 1702 | error = path_walk(pathname, &nd); |
654f562c O |
1703 | if (error) { |
1704 | if (nd.root.mnt) | |
1705 | path_put(&nd.root); | |
9b4a9b14 | 1706 | return ERR_PTR(error); |
654f562c | 1707 | } |
9b4a9b14 AV |
1708 | if (unlikely(!audit_dummy_context())) |
1709 | audit_inode(pathname, nd.path.dentry); | |
1da177e4 LT |
1710 | |
1711 | /* | |
1712 | * We have the parent and last component. First of all, check | |
1713 | * that we are not asked to creat(2) an obvious directory - that | |
1714 | * will not do. | |
1715 | */ | |
1716 | error = -EISDIR; | |
a70e65df | 1717 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1718 | goto exit_parent; |
1da177e4 | 1719 | |
8737f3a1 AV |
1720 | error = -ENFILE; |
1721 | filp = get_empty_filp(); | |
1722 | if (filp == NULL) | |
1723 | goto exit_parent; | |
1724 | nd.intent.open.file = filp; | |
1725 | nd.intent.open.flags = flag; | |
1726 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1727 | dir = nd.path.dentry; |
1728 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1729 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1730 | if (flag & O_EXCL) |
1731 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1732 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1733 | path.dentry = lookup_hash(&nd); |
1734 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1735 | |
1736 | do_last: | |
4e7506e4 AV |
1737 | error = PTR_ERR(path.dentry); |
1738 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1739 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1740 | goto exit; |
1741 | } | |
1742 | ||
a70e65df | 1743 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1744 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1745 | goto exit_mutex_unlock; |
4af4c52f OD |
1746 | } |
1747 | ||
1da177e4 | 1748 | /* Negative dentry, just create the file */ |
4e7506e4 | 1749 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1750 | /* |
1751 | * This write is needed to ensure that a | |
1752 | * ro->rw transition does not occur between | |
1753 | * the time when the file is created and when | |
1754 | * a permanent write count is taken through | |
1755 | * the 'struct file' in nameidata_to_filp(). | |
1756 | */ | |
1757 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1758 | if (error) |
4a3fd211 DH |
1759 | goto exit_mutex_unlock; |
1760 | error = __open_namei_create(&nd, &path, flag, mode); | |
1761 | if (error) { | |
1762 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1763 | goto exit; |
4a3fd211 DH |
1764 | } |
1765 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1766 | if (IS_ERR(filp)) |
1767 | ima_counts_put(&nd.path, | |
1768 | acc_mode & (MAY_READ | MAY_WRITE | | |
1769 | MAY_EXEC)); | |
4a3fd211 | 1770 | mnt_drop_write(nd.path.mnt); |
654f562c O |
1771 | if (nd.root.mnt) |
1772 | path_put(&nd.root); | |
4a3fd211 | 1773 | return filp; |
1da177e4 LT |
1774 | } |
1775 | ||
1776 | /* | |
1777 | * It already exists. | |
1778 | */ | |
1b1dcc1b | 1779 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1780 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1781 | |
1782 | error = -EEXIST; | |
1783 | if (flag & O_EXCL) | |
1784 | goto exit_dput; | |
1785 | ||
e13b210f | 1786 | if (__follow_mount(&path)) { |
1da177e4 | 1787 | error = -ELOOP; |
ba7a4c1a AV |
1788 | if (flag & O_NOFOLLOW) |
1789 | goto exit_dput; | |
1da177e4 | 1790 | } |
3e2efce0 | 1791 | |
1da177e4 | 1792 | error = -ENOENT; |
4e7506e4 | 1793 | if (!path.dentry->d_inode) |
1da177e4 | 1794 | goto exit_dput; |
acfa4380 | 1795 | if (path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1796 | goto do_link; |
1797 | ||
a70e65df | 1798 | path_to_nameidata(&path, &nd); |
1da177e4 | 1799 | error = -EISDIR; |
4e7506e4 | 1800 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1801 | goto exit; |
1802 | ok: | |
4a3fd211 DH |
1803 | /* |
1804 | * Consider: | |
1805 | * 1. may_open() truncates a file | |
1806 | * 2. a rw->ro mount transition occurs | |
1807 | * 3. nameidata_to_filp() fails due to | |
1808 | * the ro mount. | |
1809 | * That would be inconsistent, and should | |
1810 | * be avoided. Taking this mnt write here | |
1811 | * ensures that (2) can not occur. | |
1812 | */ | |
1813 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1814 | if (will_write) { | |
1815 | error = mnt_want_write(nd.path.mnt); | |
1816 | if (error) | |
1817 | goto exit; | |
1818 | } | |
3fb64190 | 1819 | error = may_open(&nd.path, acc_mode, flag); |
4a3fd211 DH |
1820 | if (error) { |
1821 | if (will_write) | |
1822 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1823 | goto exit; |
4a3fd211 DH |
1824 | } |
1825 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1826 | if (IS_ERR(filp)) |
1827 | ima_counts_put(&nd.path, | |
1828 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC)); | |
4a3fd211 DH |
1829 | /* |
1830 | * It is now safe to drop the mnt write | |
1831 | * because the filp has had a write taken | |
1832 | * on its behalf. | |
1833 | */ | |
1834 | if (will_write) | |
1835 | mnt_drop_write(nd.path.mnt); | |
654f562c O |
1836 | if (nd.root.mnt) |
1837 | path_put(&nd.root); | |
4a3fd211 | 1838 | return filp; |
1da177e4 | 1839 | |
4a3fd211 DH |
1840 | exit_mutex_unlock: |
1841 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1842 | exit_dput: |
a70e65df | 1843 | path_put_conditional(&path, &nd); |
1da177e4 | 1844 | exit: |
a70e65df CH |
1845 | if (!IS_ERR(nd.intent.open.file)) |
1846 | release_open_intent(&nd); | |
8737f3a1 | 1847 | exit_parent: |
2a737871 AV |
1848 | if (nd.root.mnt) |
1849 | path_put(&nd.root); | |
a70e65df CH |
1850 | path_put(&nd.path); |
1851 | return ERR_PTR(error); | |
1da177e4 LT |
1852 | |
1853 | do_link: | |
1854 | error = -ELOOP; | |
1855 | if (flag & O_NOFOLLOW) | |
1856 | goto exit_dput; | |
1857 | /* | |
1858 | * This is subtle. Instead of calling do_follow_link() we do the | |
1859 | * thing by hands. The reason is that this way we have zero link_count | |
1860 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1861 | * After that we have the parent and last component, i.e. | |
1862 | * we are in the same situation as after the first path_walk(). | |
1863 | * Well, almost - if the last component is normal we get its copy | |
1864 | * stored in nd->last.name and we will have to putname() it when we | |
1865 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1866 | */ | |
a70e65df CH |
1867 | nd.flags |= LOOKUP_PARENT; |
1868 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1869 | if (error) |
1870 | goto exit_dput; | |
a70e65df | 1871 | error = __do_follow_link(&path, &nd); |
de459215 KK |
1872 | if (error) { |
1873 | /* Does someone understand code flow here? Or it is only | |
1874 | * me so stupid? Anathema to whoever designed this non-sense | |
1875 | * with "intent.open". | |
1876 | */ | |
a70e65df | 1877 | release_open_intent(&nd); |
654f562c O |
1878 | if (nd.root.mnt) |
1879 | path_put(&nd.root); | |
a70e65df | 1880 | return ERR_PTR(error); |
de459215 | 1881 | } |
a70e65df CH |
1882 | nd.flags &= ~LOOKUP_PARENT; |
1883 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1884 | goto ok; |
1da177e4 | 1885 | error = -EISDIR; |
a70e65df | 1886 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1887 | goto exit; |
a70e65df CH |
1888 | if (nd.last.name[nd.last.len]) { |
1889 | __putname(nd.last.name); | |
1da177e4 LT |
1890 | goto exit; |
1891 | } | |
1892 | error = -ELOOP; | |
1893 | if (count++==32) { | |
a70e65df | 1894 | __putname(nd.last.name); |
1da177e4 LT |
1895 | goto exit; |
1896 | } | |
a70e65df | 1897 | dir = nd.path.dentry; |
1b1dcc1b | 1898 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1899 | path.dentry = lookup_hash(&nd); |
1900 | path.mnt = nd.path.mnt; | |
1901 | __putname(nd.last.name); | |
1da177e4 LT |
1902 | goto do_last; |
1903 | } | |
1904 | ||
a70e65df CH |
1905 | /** |
1906 | * filp_open - open file and return file pointer | |
1907 | * | |
1908 | * @filename: path to open | |
1909 | * @flags: open flags as per the open(2) second argument | |
1910 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1911 | * | |
1912 | * This is the helper to open a file from kernelspace if you really | |
1913 | * have to. But in generally you should not do this, so please move | |
1914 | * along, nothing to see here.. | |
1915 | */ | |
1916 | struct file *filp_open(const char *filename, int flags, int mode) | |
1917 | { | |
6e8341a1 | 1918 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1919 | } |
1920 | EXPORT_SYMBOL(filp_open); | |
1921 | ||
1da177e4 LT |
1922 | /** |
1923 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1924 | * @nd: nameidata info | |
1925 | * @is_dir: directory flag | |
1926 | * | |
1927 | * Simple function to lookup and return a dentry and create it | |
1928 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1929 | * |
4ac91378 | 1930 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1931 | */ |
1932 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1933 | { | |
c663e5d8 | 1934 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1935 | |
4ac91378 | 1936 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1937 | /* |
1938 | * Yucky last component or no last component at all? | |
1939 | * (foo/., foo/.., /////) | |
1940 | */ | |
1da177e4 LT |
1941 | if (nd->last_type != LAST_NORM) |
1942 | goto fail; | |
1943 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1944 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1945 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1946 | |
1947 | /* | |
1948 | * Do the final lookup. | |
1949 | */ | |
49705b77 | 1950 | dentry = lookup_hash(nd); |
1da177e4 LT |
1951 | if (IS_ERR(dentry)) |
1952 | goto fail; | |
c663e5d8 | 1953 | |
e9baf6e5 AV |
1954 | if (dentry->d_inode) |
1955 | goto eexist; | |
c663e5d8 CH |
1956 | /* |
1957 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1958 | * From the vfs_mknod() POV we just have a negative dentry - | |
1959 | * all is fine. Let's be bastards - you had / on the end, you've | |
1960 | * been asking for (non-existent) directory. -ENOENT for you. | |
1961 | */ | |
e9baf6e5 AV |
1962 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1963 | dput(dentry); | |
1964 | dentry = ERR_PTR(-ENOENT); | |
1965 | } | |
1da177e4 | 1966 | return dentry; |
e9baf6e5 | 1967 | eexist: |
1da177e4 | 1968 | dput(dentry); |
e9baf6e5 | 1969 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1970 | fail: |
1971 | return dentry; | |
1972 | } | |
f81a0bff | 1973 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1974 | |
1975 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1976 | { | |
a95164d9 | 1977 | int error = may_create(dir, dentry); |
1da177e4 LT |
1978 | |
1979 | if (error) | |
1980 | return error; | |
1981 | ||
1982 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1983 | return -EPERM; | |
1984 | ||
acfa4380 | 1985 | if (!dir->i_op->mknod) |
1da177e4 LT |
1986 | return -EPERM; |
1987 | ||
08ce5f16 SH |
1988 | error = devcgroup_inode_mknod(mode, dev); |
1989 | if (error) | |
1990 | return error; | |
1991 | ||
1da177e4 LT |
1992 | error = security_inode_mknod(dir, dentry, mode, dev); |
1993 | if (error) | |
1994 | return error; | |
1995 | ||
9e3509e2 | 1996 | vfs_dq_init(dir); |
1da177e4 | 1997 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1998 | if (!error) |
f38aa942 | 1999 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2000 | return error; |
2001 | } | |
2002 | ||
463c3197 DH |
2003 | static int may_mknod(mode_t mode) |
2004 | { | |
2005 | switch (mode & S_IFMT) { | |
2006 | case S_IFREG: | |
2007 | case S_IFCHR: | |
2008 | case S_IFBLK: | |
2009 | case S_IFIFO: | |
2010 | case S_IFSOCK: | |
2011 | case 0: /* zero mode translates to S_IFREG */ | |
2012 | return 0; | |
2013 | case S_IFDIR: | |
2014 | return -EPERM; | |
2015 | default: | |
2016 | return -EINVAL; | |
2017 | } | |
2018 | } | |
2019 | ||
2e4d0924 HC |
2020 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
2021 | unsigned, dev) | |
1da177e4 | 2022 | { |
2ad94ae6 AV |
2023 | int error; |
2024 | char *tmp; | |
2025 | struct dentry *dentry; | |
1da177e4 LT |
2026 | struct nameidata nd; |
2027 | ||
2028 | if (S_ISDIR(mode)) | |
2029 | return -EPERM; | |
1da177e4 | 2030 | |
2ad94ae6 | 2031 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2032 | if (error) |
2ad94ae6 AV |
2033 | return error; |
2034 | ||
1da177e4 | 2035 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2036 | if (IS_ERR(dentry)) { |
2037 | error = PTR_ERR(dentry); | |
2038 | goto out_unlock; | |
2039 | } | |
4ac91378 | 2040 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2041 | mode &= ~current_umask(); |
463c3197 DH |
2042 | error = may_mknod(mode); |
2043 | if (error) | |
2044 | goto out_dput; | |
2045 | error = mnt_want_write(nd.path.mnt); | |
2046 | if (error) | |
2047 | goto out_dput; | |
be6d3e56 KT |
2048 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2049 | if (error) | |
2050 | goto out_drop_write; | |
463c3197 | 2051 | switch (mode & S_IFMT) { |
1da177e4 | 2052 | case 0: case S_IFREG: |
4ac91378 | 2053 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2054 | break; |
2055 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2056 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2057 | new_decode_dev(dev)); |
2058 | break; | |
2059 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2060 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2061 | break; |
1da177e4 | 2062 | } |
be6d3e56 | 2063 | out_drop_write: |
463c3197 DH |
2064 | mnt_drop_write(nd.path.mnt); |
2065 | out_dput: | |
2066 | dput(dentry); | |
2067 | out_unlock: | |
4ac91378 | 2068 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2069 | path_put(&nd.path); |
1da177e4 LT |
2070 | putname(tmp); |
2071 | ||
2072 | return error; | |
2073 | } | |
2074 | ||
3480b257 | 2075 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2076 | { |
2077 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2078 | } | |
2079 | ||
1da177e4 LT |
2080 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2081 | { | |
a95164d9 | 2082 | int error = may_create(dir, dentry); |
1da177e4 LT |
2083 | |
2084 | if (error) | |
2085 | return error; | |
2086 | ||
acfa4380 | 2087 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2088 | return -EPERM; |
2089 | ||
2090 | mode &= (S_IRWXUGO|S_ISVTX); | |
2091 | error = security_inode_mkdir(dir, dentry, mode); | |
2092 | if (error) | |
2093 | return error; | |
2094 | ||
9e3509e2 | 2095 | vfs_dq_init(dir); |
1da177e4 | 2096 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2097 | if (!error) |
f38aa942 | 2098 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2099 | return error; |
2100 | } | |
2101 | ||
2e4d0924 | 2102 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2103 | { |
2104 | int error = 0; | |
2105 | char * tmp; | |
6902d925 DH |
2106 | struct dentry *dentry; |
2107 | struct nameidata nd; | |
1da177e4 | 2108 | |
2ad94ae6 AV |
2109 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2110 | if (error) | |
6902d925 | 2111 | goto out_err; |
1da177e4 | 2112 | |
6902d925 DH |
2113 | dentry = lookup_create(&nd, 1); |
2114 | error = PTR_ERR(dentry); | |
2115 | if (IS_ERR(dentry)) | |
2116 | goto out_unlock; | |
1da177e4 | 2117 | |
4ac91378 | 2118 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2119 | mode &= ~current_umask(); |
463c3197 DH |
2120 | error = mnt_want_write(nd.path.mnt); |
2121 | if (error) | |
2122 | goto out_dput; | |
be6d3e56 KT |
2123 | error = security_path_mkdir(&nd.path, dentry, mode); |
2124 | if (error) | |
2125 | goto out_drop_write; | |
4ac91378 | 2126 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2127 | out_drop_write: |
463c3197 DH |
2128 | mnt_drop_write(nd.path.mnt); |
2129 | out_dput: | |
6902d925 DH |
2130 | dput(dentry); |
2131 | out_unlock: | |
4ac91378 | 2132 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2133 | path_put(&nd.path); |
6902d925 DH |
2134 | putname(tmp); |
2135 | out_err: | |
1da177e4 LT |
2136 | return error; |
2137 | } | |
2138 | ||
3cdad428 | 2139 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2140 | { |
2141 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2142 | } | |
2143 | ||
1da177e4 LT |
2144 | /* |
2145 | * We try to drop the dentry early: we should have | |
2146 | * a usage count of 2 if we're the only user of this | |
2147 | * dentry, and if that is true (possibly after pruning | |
2148 | * the dcache), then we drop the dentry now. | |
2149 | * | |
2150 | * A low-level filesystem can, if it choses, legally | |
2151 | * do a | |
2152 | * | |
2153 | * if (!d_unhashed(dentry)) | |
2154 | * return -EBUSY; | |
2155 | * | |
2156 | * if it cannot handle the case of removing a directory | |
2157 | * that is still in use by something else.. | |
2158 | */ | |
2159 | void dentry_unhash(struct dentry *dentry) | |
2160 | { | |
2161 | dget(dentry); | |
dc168427 | 2162 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2163 | spin_lock(&dcache_lock); |
2164 | spin_lock(&dentry->d_lock); | |
2165 | if (atomic_read(&dentry->d_count) == 2) | |
2166 | __d_drop(dentry); | |
2167 | spin_unlock(&dentry->d_lock); | |
2168 | spin_unlock(&dcache_lock); | |
2169 | } | |
2170 | ||
2171 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2172 | { | |
2173 | int error = may_delete(dir, dentry, 1); | |
2174 | ||
2175 | if (error) | |
2176 | return error; | |
2177 | ||
acfa4380 | 2178 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2179 | return -EPERM; |
2180 | ||
9e3509e2 | 2181 | vfs_dq_init(dir); |
1da177e4 | 2182 | |
1b1dcc1b | 2183 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2184 | dentry_unhash(dentry); |
2185 | if (d_mountpoint(dentry)) | |
2186 | error = -EBUSY; | |
2187 | else { | |
2188 | error = security_inode_rmdir(dir, dentry); | |
2189 | if (!error) { | |
2190 | error = dir->i_op->rmdir(dir, dentry); | |
2191 | if (!error) | |
2192 | dentry->d_inode->i_flags |= S_DEAD; | |
2193 | } | |
2194 | } | |
1b1dcc1b | 2195 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2196 | if (!error) { |
1da177e4 LT |
2197 | d_delete(dentry); |
2198 | } | |
2199 | dput(dentry); | |
2200 | ||
2201 | return error; | |
2202 | } | |
2203 | ||
5590ff0d | 2204 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2205 | { |
2206 | int error = 0; | |
2207 | char * name; | |
2208 | struct dentry *dentry; | |
2209 | struct nameidata nd; | |
2210 | ||
2ad94ae6 | 2211 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2212 | if (error) |
2ad94ae6 | 2213 | return error; |
1da177e4 LT |
2214 | |
2215 | switch(nd.last_type) { | |
0612d9fb OH |
2216 | case LAST_DOTDOT: |
2217 | error = -ENOTEMPTY; | |
2218 | goto exit1; | |
2219 | case LAST_DOT: | |
2220 | error = -EINVAL; | |
2221 | goto exit1; | |
2222 | case LAST_ROOT: | |
2223 | error = -EBUSY; | |
2224 | goto exit1; | |
1da177e4 | 2225 | } |
0612d9fb OH |
2226 | |
2227 | nd.flags &= ~LOOKUP_PARENT; | |
2228 | ||
4ac91378 | 2229 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2230 | dentry = lookup_hash(&nd); |
1da177e4 | 2231 | error = PTR_ERR(dentry); |
6902d925 DH |
2232 | if (IS_ERR(dentry)) |
2233 | goto exit2; | |
0622753b DH |
2234 | error = mnt_want_write(nd.path.mnt); |
2235 | if (error) | |
2236 | goto exit3; | |
be6d3e56 KT |
2237 | error = security_path_rmdir(&nd.path, dentry); |
2238 | if (error) | |
2239 | goto exit4; | |
4ac91378 | 2240 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2241 | exit4: |
0622753b DH |
2242 | mnt_drop_write(nd.path.mnt); |
2243 | exit3: | |
6902d925 DH |
2244 | dput(dentry); |
2245 | exit2: | |
4ac91378 | 2246 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2247 | exit1: |
1d957f9b | 2248 | path_put(&nd.path); |
1da177e4 LT |
2249 | putname(name); |
2250 | return error; | |
2251 | } | |
2252 | ||
3cdad428 | 2253 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2254 | { |
2255 | return do_rmdir(AT_FDCWD, pathname); | |
2256 | } | |
2257 | ||
1da177e4 LT |
2258 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2259 | { | |
2260 | int error = may_delete(dir, dentry, 0); | |
2261 | ||
2262 | if (error) | |
2263 | return error; | |
2264 | ||
acfa4380 | 2265 | if (!dir->i_op->unlink) |
1da177e4 LT |
2266 | return -EPERM; |
2267 | ||
9e3509e2 | 2268 | vfs_dq_init(dir); |
1da177e4 | 2269 | |
1b1dcc1b | 2270 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2271 | if (d_mountpoint(dentry)) |
2272 | error = -EBUSY; | |
2273 | else { | |
2274 | error = security_inode_unlink(dir, dentry); | |
2275 | if (!error) | |
2276 | error = dir->i_op->unlink(dir, dentry); | |
2277 | } | |
1b1dcc1b | 2278 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2279 | |
2280 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2281 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2282 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2283 | d_delete(dentry); |
1da177e4 | 2284 | } |
0eeca283 | 2285 | |
1da177e4 LT |
2286 | return error; |
2287 | } | |
2288 | ||
2289 | /* | |
2290 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2291 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2292 | * writeout happening, and we don't want to prevent access to the directory |
2293 | * while waiting on the I/O. | |
2294 | */ | |
5590ff0d | 2295 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2296 | { |
2ad94ae6 AV |
2297 | int error; |
2298 | char *name; | |
1da177e4 LT |
2299 | struct dentry *dentry; |
2300 | struct nameidata nd; | |
2301 | struct inode *inode = NULL; | |
2302 | ||
2ad94ae6 | 2303 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2304 | if (error) |
2ad94ae6 AV |
2305 | return error; |
2306 | ||
1da177e4 LT |
2307 | error = -EISDIR; |
2308 | if (nd.last_type != LAST_NORM) | |
2309 | goto exit1; | |
0612d9fb OH |
2310 | |
2311 | nd.flags &= ~LOOKUP_PARENT; | |
2312 | ||
4ac91378 | 2313 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2314 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2315 | error = PTR_ERR(dentry); |
2316 | if (!IS_ERR(dentry)) { | |
2317 | /* Why not before? Because we want correct error value */ | |
2318 | if (nd.last.name[nd.last.len]) | |
2319 | goto slashes; | |
2320 | inode = dentry->d_inode; | |
2321 | if (inode) | |
2322 | atomic_inc(&inode->i_count); | |
0622753b DH |
2323 | error = mnt_want_write(nd.path.mnt); |
2324 | if (error) | |
2325 | goto exit2; | |
be6d3e56 KT |
2326 | error = security_path_unlink(&nd.path, dentry); |
2327 | if (error) | |
2328 | goto exit3; | |
4ac91378 | 2329 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2330 | exit3: |
0622753b | 2331 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2332 | exit2: |
2333 | dput(dentry); | |
2334 | } | |
4ac91378 | 2335 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2336 | if (inode) |
2337 | iput(inode); /* truncate the inode here */ | |
2338 | exit1: | |
1d957f9b | 2339 | path_put(&nd.path); |
1da177e4 LT |
2340 | putname(name); |
2341 | return error; | |
2342 | ||
2343 | slashes: | |
2344 | error = !dentry->d_inode ? -ENOENT : | |
2345 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2346 | goto exit2; | |
2347 | } | |
2348 | ||
2e4d0924 | 2349 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2350 | { |
2351 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2352 | return -EINVAL; | |
2353 | ||
2354 | if (flag & AT_REMOVEDIR) | |
2355 | return do_rmdir(dfd, pathname); | |
2356 | ||
2357 | return do_unlinkat(dfd, pathname); | |
2358 | } | |
2359 | ||
3480b257 | 2360 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2361 | { |
2362 | return do_unlinkat(AT_FDCWD, pathname); | |
2363 | } | |
2364 | ||
db2e747b | 2365 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2366 | { |
a95164d9 | 2367 | int error = may_create(dir, dentry); |
1da177e4 LT |
2368 | |
2369 | if (error) | |
2370 | return error; | |
2371 | ||
acfa4380 | 2372 | if (!dir->i_op->symlink) |
1da177e4 LT |
2373 | return -EPERM; |
2374 | ||
2375 | error = security_inode_symlink(dir, dentry, oldname); | |
2376 | if (error) | |
2377 | return error; | |
2378 | ||
9e3509e2 | 2379 | vfs_dq_init(dir); |
1da177e4 | 2380 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2381 | if (!error) |
f38aa942 | 2382 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2383 | return error; |
2384 | } | |
2385 | ||
2e4d0924 HC |
2386 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2387 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2388 | { |
2ad94ae6 AV |
2389 | int error; |
2390 | char *from; | |
2391 | char *to; | |
6902d925 DH |
2392 | struct dentry *dentry; |
2393 | struct nameidata nd; | |
1da177e4 LT |
2394 | |
2395 | from = getname(oldname); | |
2ad94ae6 | 2396 | if (IS_ERR(from)) |
1da177e4 | 2397 | return PTR_ERR(from); |
1da177e4 | 2398 | |
2ad94ae6 | 2399 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2400 | if (error) |
2ad94ae6 AV |
2401 | goto out_putname; |
2402 | ||
6902d925 DH |
2403 | dentry = lookup_create(&nd, 0); |
2404 | error = PTR_ERR(dentry); | |
2405 | if (IS_ERR(dentry)) | |
2406 | goto out_unlock; | |
2407 | ||
75c3f29d DH |
2408 | error = mnt_want_write(nd.path.mnt); |
2409 | if (error) | |
2410 | goto out_dput; | |
be6d3e56 KT |
2411 | error = security_path_symlink(&nd.path, dentry, from); |
2412 | if (error) | |
2413 | goto out_drop_write; | |
db2e747b | 2414 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2415 | out_drop_write: |
75c3f29d DH |
2416 | mnt_drop_write(nd.path.mnt); |
2417 | out_dput: | |
6902d925 DH |
2418 | dput(dentry); |
2419 | out_unlock: | |
4ac91378 | 2420 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2421 | path_put(&nd.path); |
6902d925 DH |
2422 | putname(to); |
2423 | out_putname: | |
1da177e4 LT |
2424 | putname(from); |
2425 | return error; | |
2426 | } | |
2427 | ||
3480b257 | 2428 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2429 | { |
2430 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2431 | } | |
2432 | ||
1da177e4 LT |
2433 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2434 | { | |
2435 | struct inode *inode = old_dentry->d_inode; | |
2436 | int error; | |
2437 | ||
2438 | if (!inode) | |
2439 | return -ENOENT; | |
2440 | ||
a95164d9 | 2441 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2442 | if (error) |
2443 | return error; | |
2444 | ||
2445 | if (dir->i_sb != inode->i_sb) | |
2446 | return -EXDEV; | |
2447 | ||
2448 | /* | |
2449 | * A link to an append-only or immutable file cannot be created. | |
2450 | */ | |
2451 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2452 | return -EPERM; | |
acfa4380 | 2453 | if (!dir->i_op->link) |
1da177e4 | 2454 | return -EPERM; |
7e79eedb | 2455 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2456 | return -EPERM; |
2457 | ||
2458 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2459 | if (error) | |
2460 | return error; | |
2461 | ||
7e79eedb | 2462 | mutex_lock(&inode->i_mutex); |
9e3509e2 | 2463 | vfs_dq_init(dir); |
1da177e4 | 2464 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2465 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2466 | if (!error) |
7e79eedb | 2467 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2468 | return error; |
2469 | } | |
2470 | ||
2471 | /* | |
2472 | * Hardlinks are often used in delicate situations. We avoid | |
2473 | * security-related surprises by not following symlinks on the | |
2474 | * newname. --KAB | |
2475 | * | |
2476 | * We don't follow them on the oldname either to be compatible | |
2477 | * with linux 2.0, and to avoid hard-linking to directories | |
2478 | * and other special files. --ADM | |
2479 | */ | |
2e4d0924 HC |
2480 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2481 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2482 | { |
2483 | struct dentry *new_dentry; | |
2d8f3038 AV |
2484 | struct nameidata nd; |
2485 | struct path old_path; | |
1da177e4 | 2486 | int error; |
2ad94ae6 | 2487 | char *to; |
1da177e4 | 2488 | |
45c9b11a | 2489 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2490 | return -EINVAL; |
2491 | ||
2d8f3038 AV |
2492 | error = user_path_at(olddfd, oldname, |
2493 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2494 | &old_path); | |
1da177e4 | 2495 | if (error) |
2ad94ae6 AV |
2496 | return error; |
2497 | ||
2498 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2499 | if (error) |
2500 | goto out; | |
2501 | error = -EXDEV; | |
2d8f3038 | 2502 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2503 | goto out_release; |
2504 | new_dentry = lookup_create(&nd, 0); | |
2505 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2506 | if (IS_ERR(new_dentry)) |
2507 | goto out_unlock; | |
75c3f29d DH |
2508 | error = mnt_want_write(nd.path.mnt); |
2509 | if (error) | |
2510 | goto out_dput; | |
be6d3e56 KT |
2511 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2512 | if (error) | |
2513 | goto out_drop_write; | |
2d8f3038 | 2514 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2515 | out_drop_write: |
75c3f29d DH |
2516 | mnt_drop_write(nd.path.mnt); |
2517 | out_dput: | |
6902d925 DH |
2518 | dput(new_dentry); |
2519 | out_unlock: | |
4ac91378 | 2520 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2521 | out_release: |
1d957f9b | 2522 | path_put(&nd.path); |
2ad94ae6 | 2523 | putname(to); |
1da177e4 | 2524 | out: |
2d8f3038 | 2525 | path_put(&old_path); |
1da177e4 LT |
2526 | |
2527 | return error; | |
2528 | } | |
2529 | ||
3480b257 | 2530 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2531 | { |
c04030e1 | 2532 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2533 | } |
2534 | ||
1da177e4 LT |
2535 | /* |
2536 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2537 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2538 | * Problems: | |
2539 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2540 | * b) race potential - two innocent renames can create a loop together. | |
2541 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2542 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2543 | * story. |
2544 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2545 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2546 | * whether the target exists). Solution: try to be smart with locking |
2547 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2548 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2549 | * move will be locked. Thus we can rank directories by the tree |
2550 | * (ancestors first) and rank all non-directories after them. | |
2551 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2552 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2553 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2554 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2555 | * we'd better make sure that there's no link(2) for them. | |
2556 | * d) some filesystems don't support opened-but-unlinked directories, | |
2557 | * either because of layout or because they are not ready to deal with | |
2558 | * all cases correctly. The latter will be fixed (taking this sort of | |
2559 | * stuff into VFS), but the former is not going away. Solution: the same | |
2560 | * trick as in rmdir(). | |
2561 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2562 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2563 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2564 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2565 | * locking]. |
2566 | */ | |
75c96f85 AB |
2567 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2568 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2569 | { |
2570 | int error = 0; | |
2571 | struct inode *target; | |
2572 | ||
2573 | /* | |
2574 | * If we are going to change the parent - check write permissions, | |
2575 | * we'll need to flip '..'. | |
2576 | */ | |
2577 | if (new_dir != old_dir) { | |
f419a2e3 | 2578 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2579 | if (error) |
2580 | return error; | |
2581 | } | |
2582 | ||
2583 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2584 | if (error) | |
2585 | return error; | |
2586 | ||
2587 | target = new_dentry->d_inode; | |
2588 | if (target) { | |
1b1dcc1b | 2589 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2590 | dentry_unhash(new_dentry); |
2591 | } | |
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 (target) { | |
2597 | if (!error) | |
2598 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2599 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2600 | if (d_unhashed(new_dentry)) |
2601 | d_rehash(new_dentry); | |
2602 | dput(new_dentry); | |
2603 | } | |
e31e14ec | 2604 | if (!error) |
349457cc MF |
2605 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2606 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2607 | return error; |
2608 | } | |
2609 | ||
75c96f85 AB |
2610 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2611 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2612 | { |
2613 | struct inode *target; | |
2614 | int error; | |
2615 | ||
2616 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2617 | if (error) | |
2618 | return error; | |
2619 | ||
2620 | dget(new_dentry); | |
2621 | target = new_dentry->d_inode; | |
2622 | if (target) | |
1b1dcc1b | 2623 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2624 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2625 | error = -EBUSY; | |
2626 | else | |
2627 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2628 | if (!error) { | |
349457cc | 2629 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2630 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2631 | } |
2632 | if (target) | |
1b1dcc1b | 2633 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2634 | dput(new_dentry); |
2635 | return error; | |
2636 | } | |
2637 | ||
2638 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2639 | struct inode *new_dir, struct dentry *new_dentry) | |
2640 | { | |
2641 | int error; | |
2642 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2643 | const char *old_name; |
1da177e4 LT |
2644 | |
2645 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2646 | return 0; | |
2647 | ||
2648 | error = may_delete(old_dir, old_dentry, is_dir); | |
2649 | if (error) | |
2650 | return error; | |
2651 | ||
2652 | if (!new_dentry->d_inode) | |
a95164d9 | 2653 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2654 | else |
2655 | error = may_delete(new_dir, new_dentry, is_dir); | |
2656 | if (error) | |
2657 | return error; | |
2658 | ||
acfa4380 | 2659 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2660 | return -EPERM; |
2661 | ||
9e3509e2 JK |
2662 | vfs_dq_init(old_dir); |
2663 | vfs_dq_init(new_dir); | |
1da177e4 | 2664 | |
0eeca283 RL |
2665 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2666 | ||
1da177e4 LT |
2667 | if (is_dir) |
2668 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2669 | else | |
2670 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2671 | if (!error) { | |
0eeca283 | 2672 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2673 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2674 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2675 | } |
0eeca283 RL |
2676 | fsnotify_oldname_free(old_name); |
2677 | ||
1da177e4 LT |
2678 | return error; |
2679 | } | |
2680 | ||
2e4d0924 HC |
2681 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2682 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2683 | { |
2ad94ae6 AV |
2684 | struct dentry *old_dir, *new_dir; |
2685 | struct dentry *old_dentry, *new_dentry; | |
2686 | struct dentry *trap; | |
1da177e4 | 2687 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2688 | char *from; |
2689 | char *to; | |
2690 | int error; | |
1da177e4 | 2691 | |
2ad94ae6 | 2692 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2693 | if (error) |
2694 | goto exit; | |
2695 | ||
2ad94ae6 | 2696 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2697 | if (error) |
2698 | goto exit1; | |
2699 | ||
2700 | error = -EXDEV; | |
4ac91378 | 2701 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2702 | goto exit2; |
2703 | ||
4ac91378 | 2704 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2705 | error = -EBUSY; |
2706 | if (oldnd.last_type != LAST_NORM) | |
2707 | goto exit2; | |
2708 | ||
4ac91378 | 2709 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2710 | if (newnd.last_type != LAST_NORM) |
2711 | goto exit2; | |
2712 | ||
0612d9fb OH |
2713 | oldnd.flags &= ~LOOKUP_PARENT; |
2714 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2715 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2716 | |
1da177e4 LT |
2717 | trap = lock_rename(new_dir, old_dir); |
2718 | ||
49705b77 | 2719 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2720 | error = PTR_ERR(old_dentry); |
2721 | if (IS_ERR(old_dentry)) | |
2722 | goto exit3; | |
2723 | /* source must exist */ | |
2724 | error = -ENOENT; | |
2725 | if (!old_dentry->d_inode) | |
2726 | goto exit4; | |
2727 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2728 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2729 | error = -ENOTDIR; | |
2730 | if (oldnd.last.name[oldnd.last.len]) | |
2731 | goto exit4; | |
2732 | if (newnd.last.name[newnd.last.len]) | |
2733 | goto exit4; | |
2734 | } | |
2735 | /* source should not be ancestor of target */ | |
2736 | error = -EINVAL; | |
2737 | if (old_dentry == trap) | |
2738 | goto exit4; | |
49705b77 | 2739 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2740 | error = PTR_ERR(new_dentry); |
2741 | if (IS_ERR(new_dentry)) | |
2742 | goto exit4; | |
2743 | /* target should not be an ancestor of source */ | |
2744 | error = -ENOTEMPTY; | |
2745 | if (new_dentry == trap) | |
2746 | goto exit5; | |
2747 | ||
9079b1eb DH |
2748 | error = mnt_want_write(oldnd.path.mnt); |
2749 | if (error) | |
2750 | goto exit5; | |
be6d3e56 KT |
2751 | error = security_path_rename(&oldnd.path, old_dentry, |
2752 | &newnd.path, new_dentry); | |
2753 | if (error) | |
2754 | goto exit6; | |
1da177e4 LT |
2755 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2756 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2757 | exit6: |
9079b1eb | 2758 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2759 | exit5: |
2760 | dput(new_dentry); | |
2761 | exit4: | |
2762 | dput(old_dentry); | |
2763 | exit3: | |
2764 | unlock_rename(new_dir, old_dir); | |
2765 | exit2: | |
1d957f9b | 2766 | path_put(&newnd.path); |
2ad94ae6 | 2767 | putname(to); |
1da177e4 | 2768 | exit1: |
1d957f9b | 2769 | path_put(&oldnd.path); |
1da177e4 | 2770 | putname(from); |
2ad94ae6 | 2771 | exit: |
1da177e4 LT |
2772 | return error; |
2773 | } | |
2774 | ||
a26eab24 | 2775 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2776 | { |
2777 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2778 | } | |
2779 | ||
1da177e4 LT |
2780 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2781 | { | |
2782 | int len; | |
2783 | ||
2784 | len = PTR_ERR(link); | |
2785 | if (IS_ERR(link)) | |
2786 | goto out; | |
2787 | ||
2788 | len = strlen(link); | |
2789 | if (len > (unsigned) buflen) | |
2790 | len = buflen; | |
2791 | if (copy_to_user(buffer, link, len)) | |
2792 | len = -EFAULT; | |
2793 | out: | |
2794 | return len; | |
2795 | } | |
2796 | ||
2797 | /* | |
2798 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2799 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2800 | * using) it for any given inode is up to filesystem. | |
2801 | */ | |
2802 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2803 | { | |
2804 | struct nameidata nd; | |
cc314eef | 2805 | void *cookie; |
694a1764 | 2806 | int res; |
cc314eef | 2807 | |
1da177e4 | 2808 | nd.depth = 0; |
cc314eef | 2809 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2810 | if (IS_ERR(cookie)) |
2811 | return PTR_ERR(cookie); | |
2812 | ||
2813 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2814 | if (dentry->d_inode->i_op->put_link) | |
2815 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2816 | return res; | |
1da177e4 LT |
2817 | } |
2818 | ||
2819 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2820 | { | |
2821 | return __vfs_follow_link(nd, link); | |
2822 | } | |
2823 | ||
2824 | /* get the link contents into pagecache */ | |
2825 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2826 | { | |
ebd09abb DG |
2827 | char *kaddr; |
2828 | struct page *page; | |
1da177e4 | 2829 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2830 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2831 | if (IS_ERR(page)) |
6fe6900e | 2832 | return (char*)page; |
1da177e4 | 2833 | *ppage = page; |
ebd09abb DG |
2834 | kaddr = kmap(page); |
2835 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2836 | return kaddr; | |
1da177e4 LT |
2837 | } |
2838 | ||
2839 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2840 | { | |
2841 | struct page *page = NULL; | |
2842 | char *s = page_getlink(dentry, &page); | |
2843 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2844 | if (page) { | |
2845 | kunmap(page); | |
2846 | page_cache_release(page); | |
2847 | } | |
2848 | return res; | |
2849 | } | |
2850 | ||
cc314eef | 2851 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2852 | { |
cc314eef | 2853 | struct page *page = NULL; |
1da177e4 | 2854 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2855 | return page; |
1da177e4 LT |
2856 | } |
2857 | ||
cc314eef | 2858 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2859 | { |
cc314eef LT |
2860 | struct page *page = cookie; |
2861 | ||
2862 | if (page) { | |
1da177e4 LT |
2863 | kunmap(page); |
2864 | page_cache_release(page); | |
1da177e4 LT |
2865 | } |
2866 | } | |
2867 | ||
54566b2c NP |
2868 | /* |
2869 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2870 | */ | |
2871 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2872 | { |
2873 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2874 | struct page *page; |
afddba49 | 2875 | void *fsdata; |
beb497ab | 2876 | int err; |
1da177e4 | 2877 | char *kaddr; |
54566b2c NP |
2878 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2879 | if (nofs) | |
2880 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2881 | |
7e53cac4 | 2882 | retry: |
afddba49 | 2883 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2884 | flags, &page, &fsdata); |
1da177e4 | 2885 | if (err) |
afddba49 NP |
2886 | goto fail; |
2887 | ||
1da177e4 LT |
2888 | kaddr = kmap_atomic(page, KM_USER0); |
2889 | memcpy(kaddr, symname, len-1); | |
2890 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2891 | |
2892 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2893 | page, fsdata); | |
1da177e4 LT |
2894 | if (err < 0) |
2895 | goto fail; | |
afddba49 NP |
2896 | if (err < len-1) |
2897 | goto retry; | |
2898 | ||
1da177e4 LT |
2899 | mark_inode_dirty(inode); |
2900 | return 0; | |
1da177e4 LT |
2901 | fail: |
2902 | return err; | |
2903 | } | |
2904 | ||
0adb25d2 KK |
2905 | int page_symlink(struct inode *inode, const char *symname, int len) |
2906 | { | |
2907 | return __page_symlink(inode, symname, len, | |
54566b2c | 2908 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2909 | } |
2910 | ||
92e1d5be | 2911 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2912 | .readlink = generic_readlink, |
2913 | .follow_link = page_follow_link_light, | |
2914 | .put_link = page_put_link, | |
2915 | }; | |
2916 | ||
2d8f3038 | 2917 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2918 | EXPORT_SYMBOL(follow_down); |
2919 | EXPORT_SYMBOL(follow_up); | |
2920 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2921 | EXPORT_SYMBOL(getname); | |
2922 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2923 | EXPORT_SYMBOL(lookup_one_len); |
2924 | EXPORT_SYMBOL(page_follow_link_light); | |
2925 | EXPORT_SYMBOL(page_put_link); | |
2926 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2927 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2928 | EXPORT_SYMBOL(page_symlink); |
2929 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2930 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2931 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2932 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2933 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2934 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2935 | EXPORT_SYMBOL(unlock_rename); |
2936 | EXPORT_SYMBOL(vfs_create); | |
2937 | EXPORT_SYMBOL(vfs_follow_link); | |
2938 | EXPORT_SYMBOL(vfs_link); | |
2939 | EXPORT_SYMBOL(vfs_mkdir); | |
2940 | EXPORT_SYMBOL(vfs_mknod); | |
2941 | EXPORT_SYMBOL(generic_permission); | |
2942 | EXPORT_SYMBOL(vfs_readlink); | |
2943 | EXPORT_SYMBOL(vfs_rename); | |
2944 | EXPORT_SYMBOL(vfs_rmdir); | |
2945 | EXPORT_SYMBOL(vfs_symlink); | |
2946 | EXPORT_SYMBOL(vfs_unlink); | |
2947 | EXPORT_SYMBOL(dentry_unhash); | |
2948 | EXPORT_SYMBOL(generic_readlink); |