]>
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 | ||
e81e3f4d EP |
38 | #include "internal.h" |
39 | ||
1da177e4 LT |
40 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) |
41 | ||
42 | /* [Feb-1997 T. Schoebel-Theuer] | |
43 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
44 | * were necessary because of omirr. The reason is that omirr needs | |
45 | * to know the _real_ pathname, not the user-supplied one, in case | |
46 | * of symlinks (and also when transname replacements occur). | |
47 | * | |
48 | * The new code replaces the old recursive symlink resolution with | |
49 | * an iterative one (in case of non-nested symlink chains). It does | |
50 | * this with calls to <fs>_follow_link(). | |
51 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
52 | * replaced with a single function lookup_dentry() that can handle all | |
53 | * the special cases of the former code. | |
54 | * | |
55 | * With the new dcache, the pathname is stored at each inode, at least as | |
56 | * long as the refcount of the inode is positive. As a side effect, the | |
57 | * size of the dcache depends on the inode cache and thus is dynamic. | |
58 | * | |
59 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
60 | * resolution to correspond with current state of the code. | |
61 | * | |
62 | * Note that the symlink resolution is not *completely* iterative. | |
63 | * There is still a significant amount of tail- and mid- recursion in | |
64 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
65 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
66 | * may return different results than <fs>_follow_link(). Many virtual | |
67 | * filesystems (including /proc) exhibit this behavior. | |
68 | */ | |
69 | ||
70 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
71 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
72 | * and the name already exists in form of a symlink, try to create the new | |
73 | * name indicated by the symlink. The old code always complained that the | |
74 | * name already exists, due to not following the symlink even if its target | |
75 | * is nonexistent. The new semantics affects also mknod() and link() when | |
76 | * the name is a symlink pointing to a non-existant name. | |
77 | * | |
78 | * I don't know which semantics is the right one, since I have no access | |
79 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
80 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
81 | * "old" one. Personally, I think the new semantics is much more logical. | |
82 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
83 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
84 | * and in the old Linux semantics. | |
85 | */ | |
86 | ||
87 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
88 | * semantics. See the comments in "open_namei" and "do_link" below. | |
89 | * | |
90 | * [10-Sep-98 Alan Modra] Another symlink change. | |
91 | */ | |
92 | ||
93 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
94 | * inside the path - always follow. | |
95 | * in the last component in creation/removal/renaming - never follow. | |
96 | * if LOOKUP_FOLLOW passed - follow. | |
97 | * if the pathname has trailing slashes - follow. | |
98 | * otherwise - don't follow. | |
99 | * (applied in that order). | |
100 | * | |
101 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
102 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
103 | * During the 2.4 we need to fix the userland stuff depending on it - | |
104 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
105 | * XEmacs seems to be relying on it... | |
106 | */ | |
107 | /* | |
108 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 109 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
110 | * any extra contention... |
111 | */ | |
112 | ||
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 | ||
5909ccaa LT |
172 | /* |
173 | * This does basic POSIX ACL permission checking | |
1da177e4 | 174 | */ |
5909ccaa | 175 | static int acl_permission_check(struct inode *inode, int mask, |
1da177e4 LT |
176 | int (*check_acl)(struct inode *inode, int mask)) |
177 | { | |
178 | umode_t mode = inode->i_mode; | |
179 | ||
e6305c43 AV |
180 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
181 | ||
da9592ed | 182 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
183 | mode >>= 6; |
184 | else { | |
185 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
186 | int error = check_acl(inode, mask); | |
5909ccaa | 187 | if (error != -EAGAIN) |
1da177e4 LT |
188 | return error; |
189 | } | |
190 | ||
191 | if (in_group_p(inode->i_gid)) | |
192 | mode >>= 3; | |
193 | } | |
194 | ||
195 | /* | |
196 | * If the DACs are ok we don't need any capability check. | |
197 | */ | |
e6305c43 | 198 | if ((mask & ~mode) == 0) |
1da177e4 | 199 | return 0; |
5909ccaa LT |
200 | return -EACCES; |
201 | } | |
202 | ||
203 | /** | |
204 | * generic_permission - check for access rights on a Posix-like filesystem | |
205 | * @inode: inode to check access rights for | |
206 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
207 | * @check_acl: optional callback to check for Posix ACLs | |
208 | * | |
209 | * Used to check for read/write/execute permissions on a file. | |
210 | * We use "fsuid" for this, letting us set arbitrary permissions | |
211 | * for filesystem access without changing the "normal" uids which | |
212 | * are used for other things.. | |
213 | */ | |
214 | int generic_permission(struct inode *inode, int mask, | |
215 | int (*check_acl)(struct inode *inode, int mask)) | |
216 | { | |
217 | int ret; | |
218 | ||
219 | /* | |
220 | * Do the basic POSIX ACL permission checks. | |
221 | */ | |
222 | ret = acl_permission_check(inode, mask, check_acl); | |
223 | if (ret != -EACCES) | |
224 | return ret; | |
1da177e4 | 225 | |
1da177e4 LT |
226 | /* |
227 | * Read/write DACs are always overridable. | |
228 | * Executable DACs are overridable if at least one exec bit is set. | |
229 | */ | |
f696a365 | 230 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
231 | if (capable(CAP_DAC_OVERRIDE)) |
232 | return 0; | |
233 | ||
234 | /* | |
235 | * Searching includes executable on directories, else just read. | |
236 | */ | |
237 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
238 | if (capable(CAP_DAC_READ_SEARCH)) | |
239 | return 0; | |
240 | ||
241 | return -EACCES; | |
242 | } | |
243 | ||
cb23beb5 CH |
244 | /** |
245 | * inode_permission - check for access rights to a given inode | |
246 | * @inode: inode to check permission on | |
247 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
248 | * | |
249 | * Used to check for read/write/execute permissions on an inode. | |
250 | * We use "fsuid" for this, letting us set arbitrary permissions | |
251 | * for filesystem access without changing the "normal" uids which | |
252 | * are used for other things. | |
253 | */ | |
f419a2e3 | 254 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 255 | { |
e6305c43 | 256 | int retval; |
1da177e4 LT |
257 | |
258 | if (mask & MAY_WRITE) { | |
22590e41 | 259 | umode_t mode = inode->i_mode; |
1da177e4 LT |
260 | |
261 | /* | |
262 | * Nobody gets write access to a read-only fs. | |
263 | */ | |
264 | if (IS_RDONLY(inode) && | |
265 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
266 | return -EROFS; | |
267 | ||
268 | /* | |
269 | * Nobody gets write access to an immutable file. | |
270 | */ | |
271 | if (IS_IMMUTABLE(inode)) | |
272 | return -EACCES; | |
273 | } | |
274 | ||
acfa4380 | 275 | if (inode->i_op->permission) |
b77b0646 | 276 | retval = inode->i_op->permission(inode, mask); |
f696a365 | 277 | else |
5909ccaa | 278 | retval = generic_permission(inode, mask, inode->i_op->check_acl); |
f696a365 | 279 | |
1da177e4 LT |
280 | if (retval) |
281 | return retval; | |
282 | ||
08ce5f16 SH |
283 | retval = devcgroup_inode_permission(inode, mask); |
284 | if (retval) | |
285 | return retval; | |
286 | ||
e6305c43 | 287 | return security_inode_permission(inode, |
f418b006 | 288 | mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND)); |
1da177e4 LT |
289 | } |
290 | ||
8c744fb8 CH |
291 | /** |
292 | * file_permission - check for additional access rights to a given file | |
293 | * @file: file to check access rights for | |
294 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
295 | * | |
296 | * Used to check for read/write/execute permissions on an already opened | |
297 | * file. | |
298 | * | |
299 | * Note: | |
300 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 301 | * be done using inode_permission(). |
8c744fb8 CH |
302 | */ |
303 | int file_permission(struct file *file, int mask) | |
304 | { | |
f419a2e3 | 305 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
306 | } |
307 | ||
1da177e4 LT |
308 | /* |
309 | * get_write_access() gets write permission for a file. | |
310 | * put_write_access() releases this write permission. | |
311 | * This is used for regular files. | |
312 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
313 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
314 | * can have the following values: | |
315 | * 0: no writers, no VM_DENYWRITE mappings | |
316 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
317 | * > 0: (i_writecount) users are writing to the file. | |
318 | * | |
319 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
320 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
321 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
322 | * to do the change if sign is wrong. Exclusion between them is provided by | |
323 | * the inode->i_lock spinlock. | |
324 | */ | |
325 | ||
326 | int get_write_access(struct inode * inode) | |
327 | { | |
328 | spin_lock(&inode->i_lock); | |
329 | if (atomic_read(&inode->i_writecount) < 0) { | |
330 | spin_unlock(&inode->i_lock); | |
331 | return -ETXTBSY; | |
332 | } | |
333 | atomic_inc(&inode->i_writecount); | |
334 | spin_unlock(&inode->i_lock); | |
335 | ||
336 | return 0; | |
337 | } | |
338 | ||
339 | int deny_write_access(struct file * file) | |
340 | { | |
0f7fc9e4 | 341 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
342 | |
343 | spin_lock(&inode->i_lock); | |
344 | if (atomic_read(&inode->i_writecount) > 0) { | |
345 | spin_unlock(&inode->i_lock); | |
346 | return -ETXTBSY; | |
347 | } | |
348 | atomic_dec(&inode->i_writecount); | |
349 | spin_unlock(&inode->i_lock); | |
350 | ||
351 | return 0; | |
352 | } | |
353 | ||
5dd784d0 JB |
354 | /** |
355 | * path_get - get a reference to a path | |
356 | * @path: path to get the reference to | |
357 | * | |
358 | * Given a path increment the reference count to the dentry and the vfsmount. | |
359 | */ | |
360 | void path_get(struct path *path) | |
361 | { | |
362 | mntget(path->mnt); | |
363 | dget(path->dentry); | |
364 | } | |
365 | EXPORT_SYMBOL(path_get); | |
366 | ||
1d957f9b JB |
367 | /** |
368 | * path_put - put a reference to a path | |
369 | * @path: path to put the reference to | |
370 | * | |
371 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
372 | */ | |
373 | void path_put(struct path *path) | |
1da177e4 | 374 | { |
1d957f9b JB |
375 | dput(path->dentry); |
376 | mntput(path->mnt); | |
1da177e4 | 377 | } |
1d957f9b | 378 | EXPORT_SYMBOL(path_put); |
1da177e4 | 379 | |
834f2a4a TM |
380 | /** |
381 | * release_open_intent - free up open intent resources | |
382 | * @nd: pointer to nameidata | |
383 | */ | |
384 | void release_open_intent(struct nameidata *nd) | |
385 | { | |
0f7fc9e4 | 386 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
387 | put_filp(nd->intent.open.file); |
388 | else | |
389 | fput(nd->intent.open.file); | |
390 | } | |
391 | ||
bcdc5e01 IK |
392 | static inline struct dentry * |
393 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
394 | { | |
395 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
396 | if (unlikely(status <= 0)) { | |
397 | /* | |
398 | * The dentry failed validation. | |
399 | * If d_revalidate returned 0 attempt to invalidate | |
400 | * the dentry otherwise d_revalidate is asking us | |
401 | * to return a fail status. | |
402 | */ | |
403 | if (!status) { | |
404 | if (!d_invalidate(dentry)) { | |
405 | dput(dentry); | |
406 | dentry = NULL; | |
407 | } | |
408 | } else { | |
409 | dput(dentry); | |
410 | dentry = ERR_PTR(status); | |
411 | } | |
412 | } | |
413 | return dentry; | |
414 | } | |
415 | ||
1da177e4 | 416 | /* |
b75b5086 AV |
417 | * Short-cut version of permission(), for calling on directories |
418 | * during pathname resolution. Combines parts of permission() | |
419 | * and generic_permission(), and tests ONLY for MAY_EXEC permission. | |
1da177e4 LT |
420 | * |
421 | * If appropriate, check DAC only. If not appropriate, or | |
b75b5086 | 422 | * short-cut DAC fails, then call ->permission() to do more |
1da177e4 LT |
423 | * complete permission check. |
424 | */ | |
b75b5086 | 425 | static int exec_permission(struct inode *inode) |
1da177e4 | 426 | { |
5909ccaa | 427 | int ret; |
1da177e4 | 428 | |
cb9179ea | 429 | if (inode->i_op->permission) { |
5909ccaa | 430 | ret = inode->i_op->permission(inode, MAY_EXEC); |
cb9179ea LT |
431 | if (!ret) |
432 | goto ok; | |
433 | return ret; | |
434 | } | |
5909ccaa LT |
435 | ret = acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl); |
436 | if (!ret) | |
1da177e4 LT |
437 | goto ok; |
438 | ||
f1ac9f6b | 439 | if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
440 | goto ok; |
441 | ||
5909ccaa | 442 | return ret; |
1da177e4 | 443 | ok: |
b77b0646 | 444 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
445 | } |
446 | ||
2a737871 AV |
447 | static __always_inline void set_root(struct nameidata *nd) |
448 | { | |
449 | if (!nd->root.mnt) { | |
450 | struct fs_struct *fs = current->fs; | |
451 | read_lock(&fs->lock); | |
452 | nd->root = fs->root; | |
453 | path_get(&nd->root); | |
454 | read_unlock(&fs->lock); | |
455 | } | |
456 | } | |
457 | ||
6de88d72 AV |
458 | static int link_path_walk(const char *, struct nameidata *); |
459 | ||
f1662356 | 460 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
461 | { |
462 | int res = 0; | |
463 | char *name; | |
464 | if (IS_ERR(link)) | |
465 | goto fail; | |
466 | ||
467 | if (*link == '/') { | |
2a737871 | 468 | set_root(nd); |
1d957f9b | 469 | path_put(&nd->path); |
2a737871 AV |
470 | nd->path = nd->root; |
471 | path_get(&nd->root); | |
1da177e4 | 472 | } |
b4091d5f | 473 | |
1da177e4 | 474 | res = link_path_walk(link, nd); |
1da177e4 LT |
475 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
476 | return res; | |
477 | /* | |
478 | * If it is an iterative symlinks resolution in open_namei() we | |
479 | * have to copy the last component. And all that crap because of | |
480 | * bloody create() on broken symlinks. Furrfu... | |
481 | */ | |
482 | name = __getname(); | |
483 | if (unlikely(!name)) { | |
1d957f9b | 484 | path_put(&nd->path); |
1da177e4 LT |
485 | return -ENOMEM; |
486 | } | |
487 | strcpy(name, nd->last.name); | |
488 | nd->last.name = name; | |
489 | return 0; | |
490 | fail: | |
1d957f9b | 491 | path_put(&nd->path); |
1da177e4 LT |
492 | return PTR_ERR(link); |
493 | } | |
494 | ||
1d957f9b | 495 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
496 | { |
497 | dput(path->dentry); | |
4ac91378 | 498 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
499 | mntput(path->mnt); |
500 | } | |
501 | ||
502 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
503 | { | |
4ac91378 JB |
504 | dput(nd->path.dentry); |
505 | if (nd->path.mnt != path->mnt) | |
506 | mntput(nd->path.mnt); | |
507 | nd->path.mnt = path->mnt; | |
508 | nd->path.dentry = path->dentry; | |
051d3812 IK |
509 | } |
510 | ||
f1662356 | 511 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
512 | { |
513 | int error; | |
cc314eef | 514 | void *cookie; |
cd4e91d3 | 515 | struct dentry *dentry = path->dentry; |
1da177e4 | 516 | |
d671a1cb | 517 | touch_atime(path->mnt, dentry); |
1da177e4 | 518 | nd_set_link(nd, NULL); |
cd4e91d3 | 519 | |
4ac91378 | 520 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
521 | path_to_nameidata(path, nd); |
522 | dget(dentry); | |
523 | } | |
524 | mntget(path->mnt); | |
cc314eef LT |
525 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
526 | error = PTR_ERR(cookie); | |
527 | if (!IS_ERR(cookie)) { | |
1da177e4 | 528 | char *s = nd_get_link(nd); |
cc314eef | 529 | error = 0; |
1da177e4 LT |
530 | if (s) |
531 | error = __vfs_follow_link(nd, s); | |
532 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 533 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 534 | } |
1da177e4 LT |
535 | return error; |
536 | } | |
537 | ||
538 | /* | |
539 | * This limits recursive symlink follows to 8, while | |
540 | * limiting consecutive symlinks to 40. | |
541 | * | |
542 | * Without that kind of total limit, nasty chains of consecutive | |
543 | * symlinks can cause almost arbitrarily long lookups. | |
544 | */ | |
90ebe565 | 545 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
546 | { |
547 | int err = -ELOOP; | |
548 | if (current->link_count >= MAX_NESTED_LINKS) | |
549 | goto loop; | |
550 | if (current->total_link_count >= 40) | |
551 | goto loop; | |
552 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
553 | cond_resched(); | |
90ebe565 | 554 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
555 | if (err) |
556 | goto loop; | |
557 | current->link_count++; | |
558 | current->total_link_count++; | |
559 | nd->depth++; | |
cd4e91d3 | 560 | err = __do_follow_link(path, nd); |
258fa999 | 561 | path_put(path); |
839d9f93 AV |
562 | current->link_count--; |
563 | nd->depth--; | |
1da177e4 LT |
564 | return err; |
565 | loop: | |
1d957f9b JB |
566 | path_put_conditional(path, nd); |
567 | path_put(&nd->path); | |
1da177e4 LT |
568 | return err; |
569 | } | |
570 | ||
bab77ebf | 571 | int follow_up(struct path *path) |
1da177e4 LT |
572 | { |
573 | struct vfsmount *parent; | |
574 | struct dentry *mountpoint; | |
575 | spin_lock(&vfsmount_lock); | |
bab77ebf AV |
576 | parent = path->mnt->mnt_parent; |
577 | if (parent == path->mnt) { | |
1da177e4 LT |
578 | spin_unlock(&vfsmount_lock); |
579 | return 0; | |
580 | } | |
581 | mntget(parent); | |
bab77ebf | 582 | mountpoint = dget(path->mnt->mnt_mountpoint); |
1da177e4 | 583 | spin_unlock(&vfsmount_lock); |
bab77ebf AV |
584 | dput(path->dentry); |
585 | path->dentry = mountpoint; | |
586 | mntput(path->mnt); | |
587 | path->mnt = parent; | |
1da177e4 LT |
588 | return 1; |
589 | } | |
590 | ||
591 | /* no need for dcache_lock, as serialization is taken care in | |
592 | * namespace.c | |
593 | */ | |
463ffb2e AV |
594 | static int __follow_mount(struct path *path) |
595 | { | |
596 | int res = 0; | |
597 | while (d_mountpoint(path->dentry)) { | |
1c755af4 | 598 | struct vfsmount *mounted = lookup_mnt(path); |
463ffb2e AV |
599 | if (!mounted) |
600 | break; | |
601 | dput(path->dentry); | |
602 | if (res) | |
603 | mntput(path->mnt); | |
604 | path->mnt = mounted; | |
605 | path->dentry = dget(mounted->mnt_root); | |
606 | res = 1; | |
607 | } | |
608 | return res; | |
609 | } | |
610 | ||
79ed0226 | 611 | static void follow_mount(struct path *path) |
1da177e4 | 612 | { |
79ed0226 | 613 | while (d_mountpoint(path->dentry)) { |
1c755af4 | 614 | struct vfsmount *mounted = lookup_mnt(path); |
1da177e4 LT |
615 | if (!mounted) |
616 | break; | |
79ed0226 AV |
617 | dput(path->dentry); |
618 | mntput(path->mnt); | |
619 | path->mnt = mounted; | |
620 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 | 621 | } |
1da177e4 LT |
622 | } |
623 | ||
624 | /* no need for dcache_lock, as serialization is taken care in | |
625 | * namespace.c | |
626 | */ | |
9393bd07 | 627 | int follow_down(struct path *path) |
1da177e4 LT |
628 | { |
629 | struct vfsmount *mounted; | |
630 | ||
1c755af4 | 631 | mounted = lookup_mnt(path); |
1da177e4 | 632 | if (mounted) { |
9393bd07 AV |
633 | dput(path->dentry); |
634 | mntput(path->mnt); | |
635 | path->mnt = mounted; | |
636 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
637 | return 1; |
638 | } | |
639 | return 0; | |
640 | } | |
641 | ||
f1662356 | 642 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 643 | { |
2a737871 | 644 | set_root(nd); |
e518ddb7 | 645 | |
1da177e4 LT |
646 | while(1) { |
647 | struct vfsmount *parent; | |
4ac91378 | 648 | struct dentry *old = nd->path.dentry; |
1da177e4 | 649 | |
2a737871 AV |
650 | if (nd->path.dentry == nd->root.dentry && |
651 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
652 | break; |
653 | } | |
1da177e4 | 654 | spin_lock(&dcache_lock); |
4ac91378 JB |
655 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
656 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
657 | spin_unlock(&dcache_lock); |
658 | dput(old); | |
659 | break; | |
660 | } | |
661 | spin_unlock(&dcache_lock); | |
662 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
663 | parent = nd->path.mnt->mnt_parent; |
664 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
665 | spin_unlock(&vfsmount_lock); |
666 | break; | |
667 | } | |
668 | mntget(parent); | |
4ac91378 | 669 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
670 | spin_unlock(&vfsmount_lock); |
671 | dput(old); | |
4ac91378 JB |
672 | mntput(nd->path.mnt); |
673 | nd->path.mnt = parent; | |
1da177e4 | 674 | } |
79ed0226 | 675 | follow_mount(&nd->path); |
1da177e4 LT |
676 | } |
677 | ||
1da177e4 LT |
678 | /* |
679 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
680 | * small and for now I'd prefer to have fast path as straight as possible. | |
681 | * It _is_ time-critical. | |
682 | */ | |
683 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
684 | struct path *path) | |
685 | { | |
4ac91378 | 686 | struct vfsmount *mnt = nd->path.mnt; |
6e6b1bd1 AV |
687 | struct dentry *dentry, *parent; |
688 | struct inode *dir; | |
3cac260a AV |
689 | /* |
690 | * See if the low-level filesystem might want | |
691 | * to use its own hash.. | |
692 | */ | |
693 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { | |
694 | int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, name); | |
695 | if (err < 0) | |
696 | return err; | |
697 | } | |
1da177e4 | 698 | |
3cac260a | 699 | dentry = __d_lookup(nd->path.dentry, name); |
1da177e4 LT |
700 | if (!dentry) |
701 | goto need_lookup; | |
702 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
703 | goto need_revalidate; | |
704 | done: | |
705 | path->mnt = mnt; | |
706 | path->dentry = dentry; | |
634ee701 | 707 | __follow_mount(path); |
1da177e4 LT |
708 | return 0; |
709 | ||
710 | need_lookup: | |
6e6b1bd1 AV |
711 | parent = nd->path.dentry; |
712 | dir = parent->d_inode; | |
713 | ||
714 | mutex_lock(&dir->i_mutex); | |
715 | /* | |
716 | * First re-do the cached lookup just in case it was created | |
717 | * while we waited for the directory semaphore.. | |
718 | * | |
719 | * FIXME! This could use version numbering or similar to | |
720 | * avoid unnecessary cache lookups. | |
721 | * | |
722 | * The "dcache_lock" is purely to protect the RCU list walker | |
723 | * from concurrent renames at this point (we mustn't get false | |
724 | * negatives from the RCU list walk here, unlike the optimistic | |
725 | * fast walk). | |
726 | * | |
727 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
728 | */ | |
729 | dentry = d_lookup(parent, name); | |
730 | if (!dentry) { | |
731 | struct dentry *new; | |
732 | ||
733 | /* Don't create child dentry for a dead directory. */ | |
734 | dentry = ERR_PTR(-ENOENT); | |
735 | if (IS_DEADDIR(dir)) | |
736 | goto out_unlock; | |
737 | ||
738 | new = d_alloc(parent, name); | |
739 | dentry = ERR_PTR(-ENOMEM); | |
740 | if (new) { | |
741 | dentry = dir->i_op->lookup(dir, new, nd); | |
742 | if (dentry) | |
743 | dput(new); | |
744 | else | |
745 | dentry = new; | |
746 | } | |
747 | out_unlock: | |
748 | mutex_unlock(&dir->i_mutex); | |
749 | if (IS_ERR(dentry)) | |
750 | goto fail; | |
751 | goto done; | |
752 | } | |
753 | ||
754 | /* | |
755 | * Uhhuh! Nasty case: the cache was re-populated while | |
756 | * we waited on the semaphore. Need to revalidate. | |
757 | */ | |
758 | mutex_unlock(&dir->i_mutex); | |
759 | if (dentry->d_op && dentry->d_op->d_revalidate) { | |
760 | dentry = do_revalidate(dentry, nd); | |
761 | if (!dentry) | |
762 | dentry = ERR_PTR(-ENOENT); | |
763 | } | |
1da177e4 LT |
764 | if (IS_ERR(dentry)) |
765 | goto fail; | |
766 | goto done; | |
767 | ||
768 | need_revalidate: | |
bcdc5e01 IK |
769 | dentry = do_revalidate(dentry, nd); |
770 | if (!dentry) | |
771 | goto need_lookup; | |
772 | if (IS_ERR(dentry)) | |
773 | goto fail; | |
774 | goto done; | |
1da177e4 LT |
775 | |
776 | fail: | |
777 | return PTR_ERR(dentry); | |
778 | } | |
779 | ||
780 | /* | |
781 | * Name resolution. | |
ea3834d9 PM |
782 | * This is the basic name resolution function, turning a pathname into |
783 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 784 | * |
ea3834d9 PM |
785 | * Returns 0 and nd will have valid dentry and mnt on success. |
786 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 787 | */ |
6de88d72 | 788 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
789 | { |
790 | struct path next; | |
791 | struct inode *inode; | |
792 | int err; | |
793 | unsigned int lookup_flags = nd->flags; | |
794 | ||
795 | while (*name=='/') | |
796 | name++; | |
797 | if (!*name) | |
798 | goto return_reval; | |
799 | ||
4ac91378 | 800 | inode = nd->path.dentry->d_inode; |
1da177e4 | 801 | if (nd->depth) |
f55eab82 | 802 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
803 | |
804 | /* At this point we know we have a real path component. */ | |
805 | for(;;) { | |
806 | unsigned long hash; | |
807 | struct qstr this; | |
808 | unsigned int c; | |
809 | ||
cdce5d6b | 810 | nd->flags |= LOOKUP_CONTINUE; |
b75b5086 | 811 | err = exec_permission(inode); |
1da177e4 LT |
812 | if (err) |
813 | break; | |
814 | ||
815 | this.name = name; | |
816 | c = *(const unsigned char *)name; | |
817 | ||
818 | hash = init_name_hash(); | |
819 | do { | |
820 | name++; | |
821 | hash = partial_name_hash(c, hash); | |
822 | c = *(const unsigned char *)name; | |
823 | } while (c && (c != '/')); | |
824 | this.len = name - (const char *) this.name; | |
825 | this.hash = end_name_hash(hash); | |
826 | ||
827 | /* remove trailing slashes? */ | |
828 | if (!c) | |
829 | goto last_component; | |
830 | while (*++name == '/'); | |
831 | if (!*name) | |
832 | goto last_with_slashes; | |
833 | ||
834 | /* | |
835 | * "." and ".." are special - ".." especially so because it has | |
836 | * to be able to know about the current root directory and | |
837 | * parent relationships. | |
838 | */ | |
839 | if (this.name[0] == '.') switch (this.len) { | |
840 | default: | |
841 | break; | |
842 | case 2: | |
843 | if (this.name[1] != '.') | |
844 | break; | |
58c465eb | 845 | follow_dotdot(nd); |
4ac91378 | 846 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
847 | /* fallthrough */ |
848 | case 1: | |
849 | continue; | |
850 | } | |
1da177e4 LT |
851 | /* This does the actual lookups.. */ |
852 | err = do_lookup(nd, &this, &next); | |
853 | if (err) | |
854 | break; | |
1da177e4 LT |
855 | |
856 | err = -ENOENT; | |
857 | inode = next.dentry->d_inode; | |
858 | if (!inode) | |
859 | goto out_dput; | |
1da177e4 LT |
860 | |
861 | if (inode->i_op->follow_link) { | |
90ebe565 | 862 | err = do_follow_link(&next, nd); |
1da177e4 LT |
863 | if (err) |
864 | goto return_err; | |
865 | err = -ENOENT; | |
4ac91378 | 866 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
867 | if (!inode) |
868 | break; | |
09dd17d3 MS |
869 | } else |
870 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
871 | err = -ENOTDIR; |
872 | if (!inode->i_op->lookup) | |
873 | break; | |
874 | continue; | |
875 | /* here ends the main loop */ | |
876 | ||
877 | last_with_slashes: | |
878 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
879 | last_component: | |
f55eab82 TM |
880 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
881 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
882 | if (lookup_flags & LOOKUP_PARENT) |
883 | goto lookup_parent; | |
884 | if (this.name[0] == '.') switch (this.len) { | |
885 | default: | |
886 | break; | |
887 | case 2: | |
888 | if (this.name[1] != '.') | |
889 | break; | |
58c465eb | 890 | follow_dotdot(nd); |
4ac91378 | 891 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
892 | /* fallthrough */ |
893 | case 1: | |
894 | goto return_reval; | |
895 | } | |
1da177e4 LT |
896 | err = do_lookup(nd, &this, &next); |
897 | if (err) | |
898 | break; | |
1da177e4 LT |
899 | inode = next.dentry->d_inode; |
900 | if ((lookup_flags & LOOKUP_FOLLOW) | |
acfa4380 | 901 | && inode && inode->i_op->follow_link) { |
90ebe565 | 902 | err = do_follow_link(&next, nd); |
1da177e4 LT |
903 | if (err) |
904 | goto return_err; | |
4ac91378 | 905 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
906 | } else |
907 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
908 | err = -ENOENT; |
909 | if (!inode) | |
910 | break; | |
911 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
912 | err = -ENOTDIR; | |
acfa4380 | 913 | if (!inode->i_op->lookup) |
1da177e4 LT |
914 | break; |
915 | } | |
916 | goto return_base; | |
917 | lookup_parent: | |
918 | nd->last = this; | |
919 | nd->last_type = LAST_NORM; | |
920 | if (this.name[0] != '.') | |
921 | goto return_base; | |
922 | if (this.len == 1) | |
923 | nd->last_type = LAST_DOT; | |
924 | else if (this.len == 2 && this.name[1] == '.') | |
925 | nd->last_type = LAST_DOTDOT; | |
926 | else | |
927 | goto return_base; | |
928 | return_reval: | |
929 | /* | |
930 | * We bypassed the ordinary revalidation routines. | |
931 | * We may need to check the cached dentry for staleness. | |
932 | */ | |
4ac91378 JB |
933 | if (nd->path.dentry && nd->path.dentry->d_sb && |
934 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
935 | err = -ESTALE; |
936 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
937 | if (!nd->path.dentry->d_op->d_revalidate( |
938 | nd->path.dentry, nd)) | |
1da177e4 LT |
939 | break; |
940 | } | |
941 | return_base: | |
942 | return 0; | |
943 | out_dput: | |
1d957f9b | 944 | path_put_conditional(&next, nd); |
1da177e4 LT |
945 | break; |
946 | } | |
1d957f9b | 947 | path_put(&nd->path); |
1da177e4 LT |
948 | return_err: |
949 | return err; | |
950 | } | |
951 | ||
fc9b52cd | 952 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 953 | { |
6de88d72 AV |
954 | struct path save = nd->path; |
955 | int result; | |
956 | ||
1da177e4 | 957 | current->total_link_count = 0; |
6de88d72 AV |
958 | |
959 | /* make sure the stuff we saved doesn't go away */ | |
960 | path_get(&save); | |
961 | ||
962 | result = link_path_walk(name, nd); | |
963 | if (result == -ESTALE) { | |
964 | /* nd->path had been dropped */ | |
965 | current->total_link_count = 0; | |
966 | nd->path = save; | |
967 | path_get(&nd->path); | |
968 | nd->flags |= LOOKUP_REVAL; | |
969 | result = link_path_walk(name, nd); | |
970 | } | |
971 | ||
972 | path_put(&save); | |
973 | ||
974 | return result; | |
1da177e4 LT |
975 | } |
976 | ||
9b4a9b14 | 977 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 978 | { |
ea3834d9 | 979 | int retval = 0; |
170aa3d0 UD |
980 | int fput_needed; |
981 | struct file *file; | |
1da177e4 LT |
982 | |
983 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
984 | nd->flags = flags; | |
985 | nd->depth = 0; | |
2a737871 | 986 | nd->root.mnt = NULL; |
1da177e4 | 987 | |
1da177e4 | 988 | if (*name=='/') { |
2a737871 AV |
989 | set_root(nd); |
990 | nd->path = nd->root; | |
991 | path_get(&nd->root); | |
5590ff0d | 992 | } else if (dfd == AT_FDCWD) { |
2a737871 | 993 | struct fs_struct *fs = current->fs; |
e518ddb7 | 994 | read_lock(&fs->lock); |
6ac08c39 JB |
995 | nd->path = fs->pwd; |
996 | path_get(&fs->pwd); | |
e518ddb7 | 997 | read_unlock(&fs->lock); |
5590ff0d | 998 | } else { |
5590ff0d UD |
999 | struct dentry *dentry; |
1000 | ||
1001 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1002 | retval = -EBADF; |
1003 | if (!file) | |
6d09bb62 | 1004 | goto out_fail; |
5590ff0d | 1005 | |
0f7fc9e4 | 1006 | dentry = file->f_path.dentry; |
5590ff0d | 1007 | |
170aa3d0 UD |
1008 | retval = -ENOTDIR; |
1009 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1010 | goto fput_fail; |
5590ff0d UD |
1011 | |
1012 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1013 | if (retval) |
6d09bb62 | 1014 | goto fput_fail; |
5590ff0d | 1015 | |
5dd784d0 JB |
1016 | nd->path = file->f_path; |
1017 | path_get(&file->f_path); | |
5590ff0d UD |
1018 | |
1019 | fput_light(file, fput_needed); | |
1da177e4 | 1020 | } |
9b4a9b14 | 1021 | return 0; |
2dfdd266 | 1022 | |
9b4a9b14 AV |
1023 | fput_fail: |
1024 | fput_light(file, fput_needed); | |
1025 | out_fail: | |
1026 | return retval; | |
1027 | } | |
1028 | ||
1029 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1030 | static int do_path_lookup(int dfd, const char *name, | |
1031 | unsigned int flags, struct nameidata *nd) | |
1032 | { | |
1033 | int retval = path_init(dfd, name, flags, nd); | |
1034 | if (!retval) | |
1035 | retval = path_walk(name, nd); | |
4ac91378 JB |
1036 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1037 | nd->path.dentry->d_inode)) | |
1038 | audit_inode(name, nd->path.dentry); | |
2a737871 AV |
1039 | if (nd->root.mnt) { |
1040 | path_put(&nd->root); | |
1041 | nd->root.mnt = NULL; | |
1042 | } | |
170aa3d0 | 1043 | return retval; |
1da177e4 LT |
1044 | } |
1045 | ||
fc9b52cd | 1046 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1047 | struct nameidata *nd) |
1048 | { | |
1049 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1050 | } | |
1051 | ||
d1811465 AV |
1052 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1053 | { | |
1054 | struct nameidata nd; | |
1055 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1056 | if (!res) | |
1057 | *path = nd.path; | |
1058 | return res; | |
1059 | } | |
1060 | ||
16f18200 JJS |
1061 | /** |
1062 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1063 | * @dentry: pointer to dentry of the base directory | |
1064 | * @mnt: pointer to vfs mount of the base directory | |
1065 | * @name: pointer to file name | |
1066 | * @flags: lookup flags | |
1067 | * @nd: pointer to nameidata | |
1068 | */ | |
1069 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1070 | const char *name, unsigned int flags, | |
1071 | struct nameidata *nd) | |
1072 | { | |
1073 | int retval; | |
1074 | ||
1075 | /* same as do_path_lookup */ | |
1076 | nd->last_type = LAST_ROOT; | |
1077 | nd->flags = flags; | |
1078 | nd->depth = 0; | |
1079 | ||
c8e7f449 JB |
1080 | nd->path.dentry = dentry; |
1081 | nd->path.mnt = mnt; | |
1082 | path_get(&nd->path); | |
5b857119 AV |
1083 | nd->root = nd->path; |
1084 | path_get(&nd->root); | |
16f18200 JJS |
1085 | |
1086 | retval = path_walk(name, nd); | |
4ac91378 JB |
1087 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1088 | nd->path.dentry->d_inode)) | |
1089 | audit_inode(name, nd->path.dentry); | |
16f18200 | 1090 | |
5b857119 AV |
1091 | path_put(&nd->root); |
1092 | nd->root.mnt = NULL; | |
16f18200 | 1093 | |
2a737871 | 1094 | return retval; |
16f18200 JJS |
1095 | } |
1096 | ||
eead1911 CH |
1097 | static struct dentry *__lookup_hash(struct qstr *name, |
1098 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1099 | { |
057f6c01 | 1100 | struct dentry *dentry; |
1da177e4 LT |
1101 | struct inode *inode; |
1102 | int err; | |
1103 | ||
1104 | inode = base->d_inode; | |
1da177e4 LT |
1105 | |
1106 | /* | |
1107 | * See if the low-level filesystem might want | |
1108 | * to use its own hash.. | |
1109 | */ | |
1110 | if (base->d_op && base->d_op->d_hash) { | |
1111 | err = base->d_op->d_hash(base, name); | |
1112 | dentry = ERR_PTR(err); | |
1113 | if (err < 0) | |
1114 | goto out; | |
1115 | } | |
1116 | ||
6e6b1bd1 AV |
1117 | dentry = __d_lookup(base, name); |
1118 | ||
1119 | /* lockess __d_lookup may fail due to concurrent d_move() | |
1120 | * in some unrelated directory, so try with d_lookup | |
1121 | */ | |
1122 | if (!dentry) | |
1123 | dentry = d_lookup(base, name); | |
1124 | ||
1125 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) | |
1126 | dentry = do_revalidate(dentry, nd); | |
1127 | ||
1da177e4 | 1128 | if (!dentry) { |
d70b67c8 MS |
1129 | struct dentry *new; |
1130 | ||
1131 | /* Don't create child dentry for a dead directory. */ | |
1132 | dentry = ERR_PTR(-ENOENT); | |
1133 | if (IS_DEADDIR(inode)) | |
1134 | goto out; | |
1135 | ||
1136 | new = d_alloc(base, name); | |
1da177e4 LT |
1137 | dentry = ERR_PTR(-ENOMEM); |
1138 | if (!new) | |
1139 | goto out; | |
1140 | dentry = inode->i_op->lookup(inode, new, nd); | |
1141 | if (!dentry) | |
1142 | dentry = new; | |
1143 | else | |
1144 | dput(new); | |
1145 | } | |
1146 | out: | |
1147 | return dentry; | |
1148 | } | |
1149 | ||
057f6c01 JM |
1150 | /* |
1151 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1152 | * needs parent already locked. Doesn't follow mounts. | |
1153 | * SMP-safe. | |
1154 | */ | |
eead1911 | 1155 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1156 | { |
057f6c01 JM |
1157 | int err; |
1158 | ||
b75b5086 | 1159 | err = exec_permission(nd->path.dentry->d_inode); |
057f6c01 | 1160 | if (err) |
eead1911 | 1161 | return ERR_PTR(err); |
4ac91378 | 1162 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1163 | } |
1164 | ||
eead1911 CH |
1165 | static int __lookup_one_len(const char *name, struct qstr *this, |
1166 | struct dentry *base, int len) | |
1da177e4 LT |
1167 | { |
1168 | unsigned long hash; | |
1da177e4 LT |
1169 | unsigned int c; |
1170 | ||
057f6c01 JM |
1171 | this->name = name; |
1172 | this->len = len; | |
1da177e4 | 1173 | if (!len) |
057f6c01 | 1174 | return -EACCES; |
1da177e4 LT |
1175 | |
1176 | hash = init_name_hash(); | |
1177 | while (len--) { | |
1178 | c = *(const unsigned char *)name++; | |
1179 | if (c == '/' || c == '\0') | |
057f6c01 | 1180 | return -EACCES; |
1da177e4 LT |
1181 | hash = partial_name_hash(c, hash); |
1182 | } | |
057f6c01 JM |
1183 | this->hash = end_name_hash(hash); |
1184 | return 0; | |
1185 | } | |
1da177e4 | 1186 | |
eead1911 | 1187 | /** |
a6b91919 | 1188 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1189 | * @name: pathname component to lookup |
1190 | * @base: base directory to lookup from | |
1191 | * @len: maximum length @len should be interpreted to | |
1192 | * | |
a6b91919 RD |
1193 | * Note that this routine is purely a helper for filesystem usage and should |
1194 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1195 | * nameidata argument is passed to the filesystem methods and a filesystem |
1196 | * using this helper needs to be prepared for that. | |
1197 | */ | |
057f6c01 JM |
1198 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1199 | { | |
1200 | int err; | |
1201 | struct qstr this; | |
1202 | ||
2f9092e1 DW |
1203 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1204 | ||
057f6c01 | 1205 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1206 | if (err) |
1207 | return ERR_PTR(err); | |
1208 | ||
b75b5086 | 1209 | err = exec_permission(base->d_inode); |
057f6c01 JM |
1210 | if (err) |
1211 | return ERR_PTR(err); | |
49705b77 | 1212 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1213 | } |
1214 | ||
2d8f3038 AV |
1215 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1216 | struct path *path) | |
1da177e4 | 1217 | { |
2d8f3038 | 1218 | struct nameidata nd; |
1da177e4 LT |
1219 | char *tmp = getname(name); |
1220 | int err = PTR_ERR(tmp); | |
1da177e4 | 1221 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1222 | |
1223 | BUG_ON(flags & LOOKUP_PARENT); | |
1224 | ||
1225 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1226 | putname(tmp); |
2d8f3038 AV |
1227 | if (!err) |
1228 | *path = nd.path; | |
1da177e4 LT |
1229 | } |
1230 | return err; | |
1231 | } | |
1232 | ||
2ad94ae6 AV |
1233 | static int user_path_parent(int dfd, const char __user *path, |
1234 | struct nameidata *nd, char **name) | |
1235 | { | |
1236 | char *s = getname(path); | |
1237 | int error; | |
1238 | ||
1239 | if (IS_ERR(s)) | |
1240 | return PTR_ERR(s); | |
1241 | ||
1242 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1243 | if (error) | |
1244 | putname(s); | |
1245 | else | |
1246 | *name = s; | |
1247 | ||
1248 | return error; | |
1249 | } | |
1250 | ||
1da177e4 LT |
1251 | /* |
1252 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1253 | * minimal. | |
1254 | */ | |
1255 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1256 | { | |
da9592ed DH |
1257 | uid_t fsuid = current_fsuid(); |
1258 | ||
1da177e4 LT |
1259 | if (!(dir->i_mode & S_ISVTX)) |
1260 | return 0; | |
da9592ed | 1261 | if (inode->i_uid == fsuid) |
1da177e4 | 1262 | return 0; |
da9592ed | 1263 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1264 | return 0; |
1265 | return !capable(CAP_FOWNER); | |
1266 | } | |
1267 | ||
1268 | /* | |
1269 | * Check whether we can remove a link victim from directory dir, check | |
1270 | * whether the type of victim is right. | |
1271 | * 1. We can't do it if dir is read-only (done in permission()) | |
1272 | * 2. We should have write and exec permissions on dir | |
1273 | * 3. We can't remove anything from append-only dir | |
1274 | * 4. We can't do anything with immutable dir (done in permission()) | |
1275 | * 5. If the sticky bit on dir is set we should either | |
1276 | * a. be owner of dir, or | |
1277 | * b. be owner of victim, or | |
1278 | * c. have CAP_FOWNER capability | |
1279 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1280 | * links pointing to it. | |
1281 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1282 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1283 | * 9. We can't remove a root or mountpoint. | |
1284 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1285 | * nfs_async_unlink(). | |
1286 | */ | |
858119e1 | 1287 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1288 | { |
1289 | int error; | |
1290 | ||
1291 | if (!victim->d_inode) | |
1292 | return -ENOENT; | |
1293 | ||
1294 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1295 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1296 | |
f419a2e3 | 1297 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1298 | if (error) |
1299 | return error; | |
1300 | if (IS_APPEND(dir)) | |
1301 | return -EPERM; | |
1302 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1303 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1304 | return -EPERM; |
1305 | if (isdir) { | |
1306 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1307 | return -ENOTDIR; | |
1308 | if (IS_ROOT(victim)) | |
1309 | return -EBUSY; | |
1310 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1311 | return -EISDIR; | |
1312 | if (IS_DEADDIR(dir)) | |
1313 | return -ENOENT; | |
1314 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1315 | return -EBUSY; | |
1316 | return 0; | |
1317 | } | |
1318 | ||
1319 | /* Check whether we can create an object with dentry child in directory | |
1320 | * dir. | |
1321 | * 1. We can't do it if child already exists (open has special treatment for | |
1322 | * this case, but since we are inlined it's OK) | |
1323 | * 2. We can't do it if dir is read-only (done in permission()) | |
1324 | * 3. We should have write and exec permissions on dir | |
1325 | * 4. We can't do it if dir is immutable (done in permission()) | |
1326 | */ | |
a95164d9 | 1327 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1328 | { |
1329 | if (child->d_inode) | |
1330 | return -EEXIST; | |
1331 | if (IS_DEADDIR(dir)) | |
1332 | return -ENOENT; | |
f419a2e3 | 1333 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1334 | } |
1335 | ||
1336 | /* | |
1da177e4 LT |
1337 | * O_DIRECTORY translates into forcing a directory lookup. |
1338 | */ | |
1339 | static inline int lookup_flags(unsigned int f) | |
1340 | { | |
1341 | unsigned long retval = LOOKUP_FOLLOW; | |
1342 | ||
1343 | if (f & O_NOFOLLOW) | |
1344 | retval &= ~LOOKUP_FOLLOW; | |
1345 | ||
1da177e4 LT |
1346 | if (f & O_DIRECTORY) |
1347 | retval |= LOOKUP_DIRECTORY; | |
1348 | ||
1349 | return retval; | |
1350 | } | |
1351 | ||
1352 | /* | |
1353 | * p1 and p2 should be directories on the same fs. | |
1354 | */ | |
1355 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1356 | { | |
1357 | struct dentry *p; | |
1358 | ||
1359 | if (p1 == p2) { | |
f2eace23 | 1360 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1361 | return NULL; |
1362 | } | |
1363 | ||
a11f3a05 | 1364 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1365 | |
e2761a11 OH |
1366 | p = d_ancestor(p2, p1); |
1367 | if (p) { | |
1368 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1369 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1370 | return p; | |
1da177e4 LT |
1371 | } |
1372 | ||
e2761a11 OH |
1373 | p = d_ancestor(p1, p2); |
1374 | if (p) { | |
1375 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1376 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1377 | return p; | |
1da177e4 LT |
1378 | } |
1379 | ||
f2eace23 IM |
1380 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1381 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1382 | return NULL; |
1383 | } | |
1384 | ||
1385 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1386 | { | |
1b1dcc1b | 1387 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1388 | if (p1 != p2) { |
1b1dcc1b | 1389 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1390 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1391 | } |
1392 | } | |
1393 | ||
1394 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1395 | struct nameidata *nd) | |
1396 | { | |
a95164d9 | 1397 | int error = may_create(dir, dentry); |
1da177e4 LT |
1398 | |
1399 | if (error) | |
1400 | return error; | |
1401 | ||
acfa4380 | 1402 | if (!dir->i_op->create) |
1da177e4 LT |
1403 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1404 | mode &= S_IALLUGO; | |
1405 | mode |= S_IFREG; | |
1406 | error = security_inode_create(dir, dentry, mode); | |
1407 | if (error) | |
1408 | return error; | |
9e3509e2 | 1409 | vfs_dq_init(dir); |
1da177e4 | 1410 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1411 | if (!error) |
f38aa942 | 1412 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1413 | return error; |
1414 | } | |
1415 | ||
3fb64190 | 1416 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1417 | { |
3fb64190 | 1418 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1419 | struct inode *inode = dentry->d_inode; |
1420 | int error; | |
1421 | ||
1422 | if (!inode) | |
1423 | return -ENOENT; | |
1424 | ||
c8fe8f30 CH |
1425 | switch (inode->i_mode & S_IFMT) { |
1426 | case S_IFLNK: | |
1da177e4 | 1427 | return -ELOOP; |
c8fe8f30 CH |
1428 | case S_IFDIR: |
1429 | if (acc_mode & MAY_WRITE) | |
1430 | return -EISDIR; | |
1431 | break; | |
1432 | case S_IFBLK: | |
1433 | case S_IFCHR: | |
3fb64190 | 1434 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1435 | return -EACCES; |
c8fe8f30 CH |
1436 | /*FALLTHRU*/ |
1437 | case S_IFIFO: | |
1438 | case S_IFSOCK: | |
1da177e4 | 1439 | flag &= ~O_TRUNC; |
c8fe8f30 | 1440 | break; |
4a3fd211 | 1441 | } |
b41572e9 | 1442 | |
3fb64190 | 1443 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1444 | if (error) |
1445 | return error; | |
6146f0d5 | 1446 | |
1da177e4 LT |
1447 | /* |
1448 | * An append-only file must be opened in append mode for writing. | |
1449 | */ | |
1450 | if (IS_APPEND(inode)) { | |
1451 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) | |
7715b521 | 1452 | return -EPERM; |
1da177e4 | 1453 | if (flag & O_TRUNC) |
7715b521 | 1454 | return -EPERM; |
1da177e4 LT |
1455 | } |
1456 | ||
1457 | /* O_NOATIME can only be set by the owner or superuser */ | |
7715b521 AV |
1458 | if (flag & O_NOATIME && !is_owner_or_cap(inode)) |
1459 | return -EPERM; | |
1da177e4 LT |
1460 | |
1461 | /* | |
1462 | * Ensure there are no outstanding leases on the file. | |
1463 | */ | |
1464 | error = break_lease(inode, flag); | |
1465 | if (error) | |
7715b521 | 1466 | return error; |
1da177e4 | 1467 | |
7715b521 AV |
1468 | return ima_path_check(path, acc_mode ? |
1469 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC) : | |
1470 | ACC_MODE(flag) & (MAY_READ | MAY_WRITE), | |
1471 | IMA_COUNT_UPDATE); | |
1472 | } | |
1da177e4 | 1473 | |
7715b521 AV |
1474 | static int handle_truncate(struct path *path) |
1475 | { | |
1476 | struct inode *inode = path->dentry->d_inode; | |
1477 | int error = get_write_access(inode); | |
1478 | if (error) | |
1479 | return error; | |
1480 | /* | |
1481 | * Refuse to truncate files with mandatory locks held on them. | |
1482 | */ | |
1483 | error = locks_verify_locked(inode); | |
1484 | if (!error) | |
1485 | error = security_path_truncate(path, 0, | |
1486 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN); | |
1487 | if (!error) { | |
1488 | error = do_truncate(path->dentry, 0, | |
1489 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1490 | NULL); | |
1491 | } | |
1492 | put_write_access(inode); | |
acd0c935 | 1493 | return error; |
1da177e4 LT |
1494 | } |
1495 | ||
d57999e1 DH |
1496 | /* |
1497 | * Be careful about ever adding any more callers of this | |
1498 | * function. Its flags must be in the namei format, not | |
1499 | * what get passed to sys_open(). | |
1500 | */ | |
1501 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1502 | int flag, int mode) |
1503 | { | |
1504 | int error; | |
4ac91378 | 1505 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1506 | |
1507 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1508 | mode &= ~current_umask(); |
be6d3e56 KT |
1509 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1510 | if (error) | |
1511 | goto out_unlock; | |
aab520e2 | 1512 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1513 | out_unlock: |
aab520e2 | 1514 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1515 | dput(nd->path.dentry); |
1516 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1517 | if (error) |
1518 | return error; | |
1519 | /* Don't check for write permission, don't truncate */ | |
3fb64190 | 1520 | return may_open(&nd->path, 0, flag & ~O_TRUNC); |
aab520e2 DH |
1521 | } |
1522 | ||
d57999e1 DH |
1523 | /* |
1524 | * Note that while the flag value (low two bits) for sys_open means: | |
1525 | * 00 - read-only | |
1526 | * 01 - write-only | |
1527 | * 10 - read-write | |
1528 | * 11 - special | |
1529 | * it is changed into | |
1530 | * 00 - no permissions needed | |
1531 | * 01 - read-permission | |
1532 | * 10 - write-permission | |
1533 | * 11 - read-write | |
1534 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1535 | * This is more logical, and also allows the 00 "no perm needed" | |
1536 | * to be used for symlinks (where the permissions are checked | |
1537 | * later). | |
1538 | * | |
1539 | */ | |
1540 | static inline int open_to_namei_flags(int flag) | |
1541 | { | |
1542 | if ((flag+1) & O_ACCMODE) | |
1543 | flag++; | |
1544 | return flag; | |
1545 | } | |
1546 | ||
7715b521 | 1547 | static int open_will_truncate(int flag, struct inode *inode) |
4a3fd211 DH |
1548 | { |
1549 | /* | |
1550 | * We'll never write to the fs underlying | |
1551 | * a device file. | |
1552 | */ | |
1553 | if (special_file(inode->i_mode)) | |
1554 | return 0; | |
1555 | return (flag & O_TRUNC); | |
1556 | } | |
1557 | ||
1da177e4 | 1558 | /* |
4a3fd211 DH |
1559 | * Note that the low bits of the passed in "open_flag" |
1560 | * are not the same as in the local variable "flag". See | |
1561 | * open_to_namei_flags() for more details. | |
1da177e4 | 1562 | */ |
a70e65df | 1563 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1564 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1565 | { |
4a3fd211 | 1566 | struct file *filp; |
a70e65df | 1567 | struct nameidata nd; |
6e8341a1 | 1568 | int error; |
6de88d72 | 1569 | struct path path, save; |
1da177e4 LT |
1570 | struct dentry *dir; |
1571 | int count = 0; | |
7715b521 | 1572 | int will_truncate; |
d57999e1 | 1573 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1574 | |
6b2f3d1f CH |
1575 | /* |
1576 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1577 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1578 | * always set instead of having to deal with possibly weird behaviour | |
1579 | * for malicious applications setting only __O_SYNC. | |
1580 | */ | |
1581 | if (open_flag & __O_SYNC) | |
1582 | open_flag |= O_DSYNC; | |
1583 | ||
6e8341a1 AV |
1584 | if (!acc_mode) |
1585 | acc_mode = MAY_OPEN | ACC_MODE(flag); | |
1da177e4 | 1586 | |
834f2a4a TM |
1587 | /* O_TRUNC implies we need access checks for write permissions */ |
1588 | if (flag & O_TRUNC) | |
1589 | acc_mode |= MAY_WRITE; | |
1590 | ||
1da177e4 LT |
1591 | /* Allow the LSM permission hook to distinguish append |
1592 | access from general write access. */ | |
1593 | if (flag & O_APPEND) | |
1594 | acc_mode |= MAY_APPEND; | |
1595 | ||
1da177e4 LT |
1596 | /* |
1597 | * The simplest case - just a plain lookup. | |
1598 | */ | |
1599 | if (!(flag & O_CREAT)) { | |
2dd6d1f4 AV |
1600 | filp = get_empty_filp(); |
1601 | ||
1602 | if (filp == NULL) | |
1603 | return ERR_PTR(-ENFILE); | |
1604 | nd.intent.open.file = filp; | |
1605 | nd.intent.open.flags = flag; | |
1606 | nd.intent.open.create_mode = 0; | |
1607 | error = do_path_lookup(dfd, pathname, | |
1608 | lookup_flags(flag)|LOOKUP_OPEN, &nd); | |
1609 | if (IS_ERR(nd.intent.open.file)) { | |
1610 | if (error == 0) { | |
1611 | error = PTR_ERR(nd.intent.open.file); | |
1612 | path_put(&nd.path); | |
1613 | } | |
1614 | } else if (error) | |
1615 | release_open_intent(&nd); | |
1da177e4 | 1616 | if (error) |
a70e65df | 1617 | return ERR_PTR(error); |
1da177e4 LT |
1618 | goto ok; |
1619 | } | |
1620 | ||
1621 | /* | |
1622 | * Create - we need to know the parent. | |
1623 | */ | |
9b4a9b14 | 1624 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1625 | if (error) |
a70e65df | 1626 | return ERR_PTR(error); |
9b4a9b14 | 1627 | error = path_walk(pathname, &nd); |
654f562c O |
1628 | if (error) { |
1629 | if (nd.root.mnt) | |
1630 | path_put(&nd.root); | |
9b4a9b14 | 1631 | return ERR_PTR(error); |
654f562c | 1632 | } |
9b4a9b14 AV |
1633 | if (unlikely(!audit_dummy_context())) |
1634 | audit_inode(pathname, nd.path.dentry); | |
1da177e4 LT |
1635 | |
1636 | /* | |
1637 | * We have the parent and last component. First of all, check | |
1638 | * that we are not asked to creat(2) an obvious directory - that | |
1639 | * will not do. | |
1640 | */ | |
1641 | error = -EISDIR; | |
a70e65df | 1642 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1643 | goto exit_parent; |
1da177e4 | 1644 | |
8737f3a1 AV |
1645 | error = -ENFILE; |
1646 | filp = get_empty_filp(); | |
1647 | if (filp == NULL) | |
1648 | goto exit_parent; | |
1649 | nd.intent.open.file = filp; | |
1650 | nd.intent.open.flags = flag; | |
1651 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1652 | dir = nd.path.dentry; |
1653 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1654 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1655 | if (flag & O_EXCL) |
1656 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1657 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1658 | path.dentry = lookup_hash(&nd); |
1659 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1660 | |
1661 | do_last: | |
4e7506e4 AV |
1662 | error = PTR_ERR(path.dentry); |
1663 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1664 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1665 | goto exit; |
1666 | } | |
1667 | ||
a70e65df | 1668 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1669 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1670 | goto exit_mutex_unlock; |
4af4c52f OD |
1671 | } |
1672 | ||
1da177e4 | 1673 | /* Negative dentry, just create the file */ |
4e7506e4 | 1674 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1675 | /* |
1676 | * This write is needed to ensure that a | |
1677 | * ro->rw transition does not occur between | |
1678 | * the time when the file is created and when | |
1679 | * a permanent write count is taken through | |
1680 | * the 'struct file' in nameidata_to_filp(). | |
1681 | */ | |
1682 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1683 | if (error) |
4a3fd211 DH |
1684 | goto exit_mutex_unlock; |
1685 | error = __open_namei_create(&nd, &path, flag, mode); | |
1686 | if (error) { | |
1687 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1688 | goto exit; |
4a3fd211 DH |
1689 | } |
1690 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1691 | if (IS_ERR(filp)) |
1692 | ima_counts_put(&nd.path, | |
1693 | acc_mode & (MAY_READ | MAY_WRITE | | |
1694 | MAY_EXEC)); | |
4a3fd211 | 1695 | mnt_drop_write(nd.path.mnt); |
654f562c O |
1696 | if (nd.root.mnt) |
1697 | path_put(&nd.root); | |
4a3fd211 | 1698 | return filp; |
1da177e4 LT |
1699 | } |
1700 | ||
1701 | /* | |
1702 | * It already exists. | |
1703 | */ | |
1b1dcc1b | 1704 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1705 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1706 | |
1707 | error = -EEXIST; | |
1708 | if (flag & O_EXCL) | |
1709 | goto exit_dput; | |
1710 | ||
e13b210f | 1711 | if (__follow_mount(&path)) { |
1da177e4 | 1712 | error = -ELOOP; |
ba7a4c1a AV |
1713 | if (flag & O_NOFOLLOW) |
1714 | goto exit_dput; | |
1da177e4 | 1715 | } |
3e2efce0 | 1716 | |
1da177e4 | 1717 | error = -ENOENT; |
4e7506e4 | 1718 | if (!path.dentry->d_inode) |
1da177e4 | 1719 | goto exit_dput; |
acfa4380 | 1720 | if (path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1721 | goto do_link; |
1722 | ||
a70e65df | 1723 | path_to_nameidata(&path, &nd); |
1da177e4 | 1724 | error = -EISDIR; |
4e7506e4 | 1725 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1726 | goto exit; |
1727 | ok: | |
4a3fd211 DH |
1728 | /* |
1729 | * Consider: | |
1730 | * 1. may_open() truncates a file | |
1731 | * 2. a rw->ro mount transition occurs | |
1732 | * 3. nameidata_to_filp() fails due to | |
1733 | * the ro mount. | |
1734 | * That would be inconsistent, and should | |
1735 | * be avoided. Taking this mnt write here | |
1736 | * ensures that (2) can not occur. | |
1737 | */ | |
7715b521 AV |
1738 | will_truncate = open_will_truncate(flag, nd.path.dentry->d_inode); |
1739 | if (will_truncate) { | |
4a3fd211 DH |
1740 | error = mnt_want_write(nd.path.mnt); |
1741 | if (error) | |
1742 | goto exit; | |
1743 | } | |
3fb64190 | 1744 | error = may_open(&nd.path, acc_mode, flag); |
4a3fd211 | 1745 | if (error) { |
7715b521 | 1746 | if (will_truncate) |
4a3fd211 | 1747 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 1748 | goto exit; |
4a3fd211 DH |
1749 | } |
1750 | filp = nameidata_to_filp(&nd, open_flag); | |
7715b521 | 1751 | if (IS_ERR(filp)) { |
94e5d714 MZ |
1752 | ima_counts_put(&nd.path, |
1753 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC)); | |
7715b521 AV |
1754 | if (will_truncate) |
1755 | mnt_drop_write(nd.path.mnt); | |
1756 | if (nd.root.mnt) | |
1757 | path_put(&nd.root); | |
1758 | return filp; | |
1759 | } | |
1760 | ||
1761 | if (acc_mode & MAY_WRITE) | |
1762 | vfs_dq_init(nd.path.dentry->d_inode); | |
1763 | ||
1764 | if (will_truncate) { | |
1765 | error = handle_truncate(&nd.path); | |
1766 | if (error) { | |
1767 | mnt_drop_write(nd.path.mnt); | |
1768 | fput(filp); | |
1769 | if (nd.root.mnt) | |
1770 | path_put(&nd.root); | |
1771 | return ERR_PTR(error); | |
1772 | } | |
1773 | } | |
4a3fd211 DH |
1774 | /* |
1775 | * It is now safe to drop the mnt write | |
1776 | * because the filp has had a write taken | |
1777 | * on its behalf. | |
1778 | */ | |
7715b521 | 1779 | if (will_truncate) |
4a3fd211 | 1780 | mnt_drop_write(nd.path.mnt); |
654f562c O |
1781 | if (nd.root.mnt) |
1782 | path_put(&nd.root); | |
4a3fd211 | 1783 | return filp; |
1da177e4 | 1784 | |
4a3fd211 DH |
1785 | exit_mutex_unlock: |
1786 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1787 | exit_dput: |
a70e65df | 1788 | path_put_conditional(&path, &nd); |
1da177e4 | 1789 | exit: |
a70e65df CH |
1790 | if (!IS_ERR(nd.intent.open.file)) |
1791 | release_open_intent(&nd); | |
8737f3a1 | 1792 | exit_parent: |
2a737871 AV |
1793 | if (nd.root.mnt) |
1794 | path_put(&nd.root); | |
a70e65df CH |
1795 | path_put(&nd.path); |
1796 | return ERR_PTR(error); | |
1da177e4 LT |
1797 | |
1798 | do_link: | |
1799 | error = -ELOOP; | |
1800 | if (flag & O_NOFOLLOW) | |
1801 | goto exit_dput; | |
1802 | /* | |
1803 | * This is subtle. Instead of calling do_follow_link() we do the | |
1804 | * thing by hands. The reason is that this way we have zero link_count | |
1805 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1806 | * After that we have the parent and last component, i.e. | |
1807 | * we are in the same situation as after the first path_walk(). | |
1808 | * Well, almost - if the last component is normal we get its copy | |
1809 | * stored in nd->last.name and we will have to putname() it when we | |
1810 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1811 | */ | |
a70e65df CH |
1812 | nd.flags |= LOOKUP_PARENT; |
1813 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1814 | if (error) |
1815 | goto exit_dput; | |
6de88d72 AV |
1816 | save = nd.path; |
1817 | path_get(&save); | |
a70e65df | 1818 | error = __do_follow_link(&path, &nd); |
6de88d72 AV |
1819 | if (error == -ESTALE) { |
1820 | /* nd.path had been dropped */ | |
1821 | nd.path = save; | |
1822 | path_get(&nd.path); | |
1823 | nd.flags |= LOOKUP_REVAL; | |
1824 | error = __do_follow_link(&path, &nd); | |
1825 | } | |
1826 | path_put(&save); | |
258fa999 | 1827 | path_put(&path); |
de459215 KK |
1828 | if (error) { |
1829 | /* Does someone understand code flow here? Or it is only | |
1830 | * me so stupid? Anathema to whoever designed this non-sense | |
1831 | * with "intent.open". | |
1832 | */ | |
a70e65df | 1833 | release_open_intent(&nd); |
654f562c O |
1834 | if (nd.root.mnt) |
1835 | path_put(&nd.root); | |
a70e65df | 1836 | return ERR_PTR(error); |
de459215 | 1837 | } |
a70e65df CH |
1838 | nd.flags &= ~LOOKUP_PARENT; |
1839 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1840 | goto ok; |
1da177e4 | 1841 | error = -EISDIR; |
a70e65df | 1842 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1843 | goto exit; |
a70e65df CH |
1844 | if (nd.last.name[nd.last.len]) { |
1845 | __putname(nd.last.name); | |
1da177e4 LT |
1846 | goto exit; |
1847 | } | |
1848 | error = -ELOOP; | |
1849 | if (count++==32) { | |
a70e65df | 1850 | __putname(nd.last.name); |
1da177e4 LT |
1851 | goto exit; |
1852 | } | |
a70e65df | 1853 | dir = nd.path.dentry; |
1b1dcc1b | 1854 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1855 | path.dentry = lookup_hash(&nd); |
1856 | path.mnt = nd.path.mnt; | |
1857 | __putname(nd.last.name); | |
1da177e4 LT |
1858 | goto do_last; |
1859 | } | |
1860 | ||
a70e65df CH |
1861 | /** |
1862 | * filp_open - open file and return file pointer | |
1863 | * | |
1864 | * @filename: path to open | |
1865 | * @flags: open flags as per the open(2) second argument | |
1866 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1867 | * | |
1868 | * This is the helper to open a file from kernelspace if you really | |
1869 | * have to. But in generally you should not do this, so please move | |
1870 | * along, nothing to see here.. | |
1871 | */ | |
1872 | struct file *filp_open(const char *filename, int flags, int mode) | |
1873 | { | |
6e8341a1 | 1874 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1875 | } |
1876 | EXPORT_SYMBOL(filp_open); | |
1877 | ||
1da177e4 LT |
1878 | /** |
1879 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1880 | * @nd: nameidata info | |
1881 | * @is_dir: directory flag | |
1882 | * | |
1883 | * Simple function to lookup and return a dentry and create it | |
1884 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1885 | * |
4ac91378 | 1886 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1887 | */ |
1888 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1889 | { | |
c663e5d8 | 1890 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1891 | |
4ac91378 | 1892 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1893 | /* |
1894 | * Yucky last component or no last component at all? | |
1895 | * (foo/., foo/.., /////) | |
1896 | */ | |
1da177e4 LT |
1897 | if (nd->last_type != LAST_NORM) |
1898 | goto fail; | |
1899 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1900 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1901 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1902 | |
1903 | /* | |
1904 | * Do the final lookup. | |
1905 | */ | |
49705b77 | 1906 | dentry = lookup_hash(nd); |
1da177e4 LT |
1907 | if (IS_ERR(dentry)) |
1908 | goto fail; | |
c663e5d8 | 1909 | |
e9baf6e5 AV |
1910 | if (dentry->d_inode) |
1911 | goto eexist; | |
c663e5d8 CH |
1912 | /* |
1913 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1914 | * From the vfs_mknod() POV we just have a negative dentry - | |
1915 | * all is fine. Let's be bastards - you had / on the end, you've | |
1916 | * been asking for (non-existent) directory. -ENOENT for you. | |
1917 | */ | |
e9baf6e5 AV |
1918 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1919 | dput(dentry); | |
1920 | dentry = ERR_PTR(-ENOENT); | |
1921 | } | |
1da177e4 | 1922 | return dentry; |
e9baf6e5 | 1923 | eexist: |
1da177e4 | 1924 | dput(dentry); |
e9baf6e5 | 1925 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1926 | fail: |
1927 | return dentry; | |
1928 | } | |
f81a0bff | 1929 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1930 | |
1931 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1932 | { | |
a95164d9 | 1933 | int error = may_create(dir, dentry); |
1da177e4 LT |
1934 | |
1935 | if (error) | |
1936 | return error; | |
1937 | ||
1938 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1939 | return -EPERM; | |
1940 | ||
acfa4380 | 1941 | if (!dir->i_op->mknod) |
1da177e4 LT |
1942 | return -EPERM; |
1943 | ||
08ce5f16 SH |
1944 | error = devcgroup_inode_mknod(mode, dev); |
1945 | if (error) | |
1946 | return error; | |
1947 | ||
1da177e4 LT |
1948 | error = security_inode_mknod(dir, dentry, mode, dev); |
1949 | if (error) | |
1950 | return error; | |
1951 | ||
9e3509e2 | 1952 | vfs_dq_init(dir); |
1da177e4 | 1953 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1954 | if (!error) |
f38aa942 | 1955 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1956 | return error; |
1957 | } | |
1958 | ||
463c3197 DH |
1959 | static int may_mknod(mode_t mode) |
1960 | { | |
1961 | switch (mode & S_IFMT) { | |
1962 | case S_IFREG: | |
1963 | case S_IFCHR: | |
1964 | case S_IFBLK: | |
1965 | case S_IFIFO: | |
1966 | case S_IFSOCK: | |
1967 | case 0: /* zero mode translates to S_IFREG */ | |
1968 | return 0; | |
1969 | case S_IFDIR: | |
1970 | return -EPERM; | |
1971 | default: | |
1972 | return -EINVAL; | |
1973 | } | |
1974 | } | |
1975 | ||
2e4d0924 HC |
1976 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
1977 | unsigned, dev) | |
1da177e4 | 1978 | { |
2ad94ae6 AV |
1979 | int error; |
1980 | char *tmp; | |
1981 | struct dentry *dentry; | |
1da177e4 LT |
1982 | struct nameidata nd; |
1983 | ||
1984 | if (S_ISDIR(mode)) | |
1985 | return -EPERM; | |
1da177e4 | 1986 | |
2ad94ae6 | 1987 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 1988 | if (error) |
2ad94ae6 AV |
1989 | return error; |
1990 | ||
1da177e4 | 1991 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
1992 | if (IS_ERR(dentry)) { |
1993 | error = PTR_ERR(dentry); | |
1994 | goto out_unlock; | |
1995 | } | |
4ac91378 | 1996 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 1997 | mode &= ~current_umask(); |
463c3197 DH |
1998 | error = may_mknod(mode); |
1999 | if (error) | |
2000 | goto out_dput; | |
2001 | error = mnt_want_write(nd.path.mnt); | |
2002 | if (error) | |
2003 | goto out_dput; | |
be6d3e56 KT |
2004 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2005 | if (error) | |
2006 | goto out_drop_write; | |
463c3197 | 2007 | switch (mode & S_IFMT) { |
1da177e4 | 2008 | case 0: case S_IFREG: |
4ac91378 | 2009 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2010 | break; |
2011 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2012 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2013 | new_decode_dev(dev)); |
2014 | break; | |
2015 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2016 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2017 | break; |
1da177e4 | 2018 | } |
be6d3e56 | 2019 | out_drop_write: |
463c3197 DH |
2020 | mnt_drop_write(nd.path.mnt); |
2021 | out_dput: | |
2022 | dput(dentry); | |
2023 | out_unlock: | |
4ac91378 | 2024 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2025 | path_put(&nd.path); |
1da177e4 LT |
2026 | putname(tmp); |
2027 | ||
2028 | return error; | |
2029 | } | |
2030 | ||
3480b257 | 2031 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2032 | { |
2033 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2034 | } | |
2035 | ||
1da177e4 LT |
2036 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2037 | { | |
a95164d9 | 2038 | int error = may_create(dir, dentry); |
1da177e4 LT |
2039 | |
2040 | if (error) | |
2041 | return error; | |
2042 | ||
acfa4380 | 2043 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2044 | return -EPERM; |
2045 | ||
2046 | mode &= (S_IRWXUGO|S_ISVTX); | |
2047 | error = security_inode_mkdir(dir, dentry, mode); | |
2048 | if (error) | |
2049 | return error; | |
2050 | ||
9e3509e2 | 2051 | vfs_dq_init(dir); |
1da177e4 | 2052 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2053 | if (!error) |
f38aa942 | 2054 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2055 | return error; |
2056 | } | |
2057 | ||
2e4d0924 | 2058 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2059 | { |
2060 | int error = 0; | |
2061 | char * tmp; | |
6902d925 DH |
2062 | struct dentry *dentry; |
2063 | struct nameidata nd; | |
1da177e4 | 2064 | |
2ad94ae6 AV |
2065 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2066 | if (error) | |
6902d925 | 2067 | goto out_err; |
1da177e4 | 2068 | |
6902d925 DH |
2069 | dentry = lookup_create(&nd, 1); |
2070 | error = PTR_ERR(dentry); | |
2071 | if (IS_ERR(dentry)) | |
2072 | goto out_unlock; | |
1da177e4 | 2073 | |
4ac91378 | 2074 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2075 | mode &= ~current_umask(); |
463c3197 DH |
2076 | error = mnt_want_write(nd.path.mnt); |
2077 | if (error) | |
2078 | goto out_dput; | |
be6d3e56 KT |
2079 | error = security_path_mkdir(&nd.path, dentry, mode); |
2080 | if (error) | |
2081 | goto out_drop_write; | |
4ac91378 | 2082 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2083 | out_drop_write: |
463c3197 DH |
2084 | mnt_drop_write(nd.path.mnt); |
2085 | out_dput: | |
6902d925 DH |
2086 | dput(dentry); |
2087 | out_unlock: | |
4ac91378 | 2088 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2089 | path_put(&nd.path); |
6902d925 DH |
2090 | putname(tmp); |
2091 | out_err: | |
1da177e4 LT |
2092 | return error; |
2093 | } | |
2094 | ||
3cdad428 | 2095 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2096 | { |
2097 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2098 | } | |
2099 | ||
1da177e4 LT |
2100 | /* |
2101 | * We try to drop the dentry early: we should have | |
2102 | * a usage count of 2 if we're the only user of this | |
2103 | * dentry, and if that is true (possibly after pruning | |
2104 | * the dcache), then we drop the dentry now. | |
2105 | * | |
2106 | * A low-level filesystem can, if it choses, legally | |
2107 | * do a | |
2108 | * | |
2109 | * if (!d_unhashed(dentry)) | |
2110 | * return -EBUSY; | |
2111 | * | |
2112 | * if it cannot handle the case of removing a directory | |
2113 | * that is still in use by something else.. | |
2114 | */ | |
2115 | void dentry_unhash(struct dentry *dentry) | |
2116 | { | |
2117 | dget(dentry); | |
dc168427 | 2118 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2119 | spin_lock(&dcache_lock); |
2120 | spin_lock(&dentry->d_lock); | |
2121 | if (atomic_read(&dentry->d_count) == 2) | |
2122 | __d_drop(dentry); | |
2123 | spin_unlock(&dentry->d_lock); | |
2124 | spin_unlock(&dcache_lock); | |
2125 | } | |
2126 | ||
2127 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2128 | { | |
2129 | int error = may_delete(dir, dentry, 1); | |
2130 | ||
2131 | if (error) | |
2132 | return error; | |
2133 | ||
acfa4380 | 2134 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2135 | return -EPERM; |
2136 | ||
9e3509e2 | 2137 | vfs_dq_init(dir); |
1da177e4 | 2138 | |
1b1dcc1b | 2139 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2140 | dentry_unhash(dentry); |
2141 | if (d_mountpoint(dentry)) | |
2142 | error = -EBUSY; | |
2143 | else { | |
2144 | error = security_inode_rmdir(dir, dentry); | |
2145 | if (!error) { | |
2146 | error = dir->i_op->rmdir(dir, dentry); | |
2147 | if (!error) | |
2148 | dentry->d_inode->i_flags |= S_DEAD; | |
2149 | } | |
2150 | } | |
1b1dcc1b | 2151 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2152 | if (!error) { |
1da177e4 LT |
2153 | d_delete(dentry); |
2154 | } | |
2155 | dput(dentry); | |
2156 | ||
2157 | return error; | |
2158 | } | |
2159 | ||
5590ff0d | 2160 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2161 | { |
2162 | int error = 0; | |
2163 | char * name; | |
2164 | struct dentry *dentry; | |
2165 | struct nameidata nd; | |
2166 | ||
2ad94ae6 | 2167 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2168 | if (error) |
2ad94ae6 | 2169 | return error; |
1da177e4 LT |
2170 | |
2171 | switch(nd.last_type) { | |
0612d9fb OH |
2172 | case LAST_DOTDOT: |
2173 | error = -ENOTEMPTY; | |
2174 | goto exit1; | |
2175 | case LAST_DOT: | |
2176 | error = -EINVAL; | |
2177 | goto exit1; | |
2178 | case LAST_ROOT: | |
2179 | error = -EBUSY; | |
2180 | goto exit1; | |
1da177e4 | 2181 | } |
0612d9fb OH |
2182 | |
2183 | nd.flags &= ~LOOKUP_PARENT; | |
2184 | ||
4ac91378 | 2185 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2186 | dentry = lookup_hash(&nd); |
1da177e4 | 2187 | error = PTR_ERR(dentry); |
6902d925 DH |
2188 | if (IS_ERR(dentry)) |
2189 | goto exit2; | |
0622753b DH |
2190 | error = mnt_want_write(nd.path.mnt); |
2191 | if (error) | |
2192 | goto exit3; | |
be6d3e56 KT |
2193 | error = security_path_rmdir(&nd.path, dentry); |
2194 | if (error) | |
2195 | goto exit4; | |
4ac91378 | 2196 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2197 | exit4: |
0622753b DH |
2198 | mnt_drop_write(nd.path.mnt); |
2199 | exit3: | |
6902d925 DH |
2200 | dput(dentry); |
2201 | exit2: | |
4ac91378 | 2202 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2203 | exit1: |
1d957f9b | 2204 | path_put(&nd.path); |
1da177e4 LT |
2205 | putname(name); |
2206 | return error; | |
2207 | } | |
2208 | ||
3cdad428 | 2209 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2210 | { |
2211 | return do_rmdir(AT_FDCWD, pathname); | |
2212 | } | |
2213 | ||
1da177e4 LT |
2214 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2215 | { | |
2216 | int error = may_delete(dir, dentry, 0); | |
2217 | ||
2218 | if (error) | |
2219 | return error; | |
2220 | ||
acfa4380 | 2221 | if (!dir->i_op->unlink) |
1da177e4 LT |
2222 | return -EPERM; |
2223 | ||
9e3509e2 | 2224 | vfs_dq_init(dir); |
1da177e4 | 2225 | |
1b1dcc1b | 2226 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2227 | if (d_mountpoint(dentry)) |
2228 | error = -EBUSY; | |
2229 | else { | |
2230 | error = security_inode_unlink(dir, dentry); | |
2231 | if (!error) | |
2232 | error = dir->i_op->unlink(dir, dentry); | |
2233 | } | |
1b1dcc1b | 2234 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2235 | |
2236 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2237 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2238 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2239 | d_delete(dentry); |
1da177e4 | 2240 | } |
0eeca283 | 2241 | |
1da177e4 LT |
2242 | return error; |
2243 | } | |
2244 | ||
2245 | /* | |
2246 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2247 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2248 | * writeout happening, and we don't want to prevent access to the directory |
2249 | * while waiting on the I/O. | |
2250 | */ | |
5590ff0d | 2251 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2252 | { |
2ad94ae6 AV |
2253 | int error; |
2254 | char *name; | |
1da177e4 LT |
2255 | struct dentry *dentry; |
2256 | struct nameidata nd; | |
2257 | struct inode *inode = NULL; | |
2258 | ||
2ad94ae6 | 2259 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2260 | if (error) |
2ad94ae6 AV |
2261 | return error; |
2262 | ||
1da177e4 LT |
2263 | error = -EISDIR; |
2264 | if (nd.last_type != LAST_NORM) | |
2265 | goto exit1; | |
0612d9fb OH |
2266 | |
2267 | nd.flags &= ~LOOKUP_PARENT; | |
2268 | ||
4ac91378 | 2269 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2270 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2271 | error = PTR_ERR(dentry); |
2272 | if (!IS_ERR(dentry)) { | |
2273 | /* Why not before? Because we want correct error value */ | |
2274 | if (nd.last.name[nd.last.len]) | |
2275 | goto slashes; | |
2276 | inode = dentry->d_inode; | |
2277 | if (inode) | |
2278 | atomic_inc(&inode->i_count); | |
0622753b DH |
2279 | error = mnt_want_write(nd.path.mnt); |
2280 | if (error) | |
2281 | goto exit2; | |
be6d3e56 KT |
2282 | error = security_path_unlink(&nd.path, dentry); |
2283 | if (error) | |
2284 | goto exit3; | |
4ac91378 | 2285 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2286 | exit3: |
0622753b | 2287 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2288 | exit2: |
2289 | dput(dentry); | |
2290 | } | |
4ac91378 | 2291 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2292 | if (inode) |
2293 | iput(inode); /* truncate the inode here */ | |
2294 | exit1: | |
1d957f9b | 2295 | path_put(&nd.path); |
1da177e4 LT |
2296 | putname(name); |
2297 | return error; | |
2298 | ||
2299 | slashes: | |
2300 | error = !dentry->d_inode ? -ENOENT : | |
2301 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2302 | goto exit2; | |
2303 | } | |
2304 | ||
2e4d0924 | 2305 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2306 | { |
2307 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2308 | return -EINVAL; | |
2309 | ||
2310 | if (flag & AT_REMOVEDIR) | |
2311 | return do_rmdir(dfd, pathname); | |
2312 | ||
2313 | return do_unlinkat(dfd, pathname); | |
2314 | } | |
2315 | ||
3480b257 | 2316 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2317 | { |
2318 | return do_unlinkat(AT_FDCWD, pathname); | |
2319 | } | |
2320 | ||
db2e747b | 2321 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2322 | { |
a95164d9 | 2323 | int error = may_create(dir, dentry); |
1da177e4 LT |
2324 | |
2325 | if (error) | |
2326 | return error; | |
2327 | ||
acfa4380 | 2328 | if (!dir->i_op->symlink) |
1da177e4 LT |
2329 | return -EPERM; |
2330 | ||
2331 | error = security_inode_symlink(dir, dentry, oldname); | |
2332 | if (error) | |
2333 | return error; | |
2334 | ||
9e3509e2 | 2335 | vfs_dq_init(dir); |
1da177e4 | 2336 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2337 | if (!error) |
f38aa942 | 2338 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2339 | return error; |
2340 | } | |
2341 | ||
2e4d0924 HC |
2342 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2343 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2344 | { |
2ad94ae6 AV |
2345 | int error; |
2346 | char *from; | |
2347 | char *to; | |
6902d925 DH |
2348 | struct dentry *dentry; |
2349 | struct nameidata nd; | |
1da177e4 LT |
2350 | |
2351 | from = getname(oldname); | |
2ad94ae6 | 2352 | if (IS_ERR(from)) |
1da177e4 | 2353 | return PTR_ERR(from); |
1da177e4 | 2354 | |
2ad94ae6 | 2355 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2356 | if (error) |
2ad94ae6 AV |
2357 | goto out_putname; |
2358 | ||
6902d925 DH |
2359 | dentry = lookup_create(&nd, 0); |
2360 | error = PTR_ERR(dentry); | |
2361 | if (IS_ERR(dentry)) | |
2362 | goto out_unlock; | |
2363 | ||
75c3f29d DH |
2364 | error = mnt_want_write(nd.path.mnt); |
2365 | if (error) | |
2366 | goto out_dput; | |
be6d3e56 KT |
2367 | error = security_path_symlink(&nd.path, dentry, from); |
2368 | if (error) | |
2369 | goto out_drop_write; | |
db2e747b | 2370 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2371 | out_drop_write: |
75c3f29d DH |
2372 | mnt_drop_write(nd.path.mnt); |
2373 | out_dput: | |
6902d925 DH |
2374 | dput(dentry); |
2375 | out_unlock: | |
4ac91378 | 2376 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2377 | path_put(&nd.path); |
6902d925 DH |
2378 | putname(to); |
2379 | out_putname: | |
1da177e4 LT |
2380 | putname(from); |
2381 | return error; | |
2382 | } | |
2383 | ||
3480b257 | 2384 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2385 | { |
2386 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2387 | } | |
2388 | ||
1da177e4 LT |
2389 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2390 | { | |
2391 | struct inode *inode = old_dentry->d_inode; | |
2392 | int error; | |
2393 | ||
2394 | if (!inode) | |
2395 | return -ENOENT; | |
2396 | ||
a95164d9 | 2397 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2398 | if (error) |
2399 | return error; | |
2400 | ||
2401 | if (dir->i_sb != inode->i_sb) | |
2402 | return -EXDEV; | |
2403 | ||
2404 | /* | |
2405 | * A link to an append-only or immutable file cannot be created. | |
2406 | */ | |
2407 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2408 | return -EPERM; | |
acfa4380 | 2409 | if (!dir->i_op->link) |
1da177e4 | 2410 | return -EPERM; |
7e79eedb | 2411 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2412 | return -EPERM; |
2413 | ||
2414 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2415 | if (error) | |
2416 | return error; | |
2417 | ||
7e79eedb | 2418 | mutex_lock(&inode->i_mutex); |
9e3509e2 | 2419 | vfs_dq_init(dir); |
1da177e4 | 2420 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2421 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2422 | if (!error) |
7e79eedb | 2423 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2424 | return error; |
2425 | } | |
2426 | ||
2427 | /* | |
2428 | * Hardlinks are often used in delicate situations. We avoid | |
2429 | * security-related surprises by not following symlinks on the | |
2430 | * newname. --KAB | |
2431 | * | |
2432 | * We don't follow them on the oldname either to be compatible | |
2433 | * with linux 2.0, and to avoid hard-linking to directories | |
2434 | * and other special files. --ADM | |
2435 | */ | |
2e4d0924 HC |
2436 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2437 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2438 | { |
2439 | struct dentry *new_dentry; | |
2d8f3038 AV |
2440 | struct nameidata nd; |
2441 | struct path old_path; | |
1da177e4 | 2442 | int error; |
2ad94ae6 | 2443 | char *to; |
1da177e4 | 2444 | |
45c9b11a | 2445 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2446 | return -EINVAL; |
2447 | ||
2d8f3038 AV |
2448 | error = user_path_at(olddfd, oldname, |
2449 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2450 | &old_path); | |
1da177e4 | 2451 | if (error) |
2ad94ae6 AV |
2452 | return error; |
2453 | ||
2454 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2455 | if (error) |
2456 | goto out; | |
2457 | error = -EXDEV; | |
2d8f3038 | 2458 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2459 | goto out_release; |
2460 | new_dentry = lookup_create(&nd, 0); | |
2461 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2462 | if (IS_ERR(new_dentry)) |
2463 | goto out_unlock; | |
75c3f29d DH |
2464 | error = mnt_want_write(nd.path.mnt); |
2465 | if (error) | |
2466 | goto out_dput; | |
be6d3e56 KT |
2467 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2468 | if (error) | |
2469 | goto out_drop_write; | |
2d8f3038 | 2470 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2471 | out_drop_write: |
75c3f29d DH |
2472 | mnt_drop_write(nd.path.mnt); |
2473 | out_dput: | |
6902d925 DH |
2474 | dput(new_dentry); |
2475 | out_unlock: | |
4ac91378 | 2476 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2477 | out_release: |
1d957f9b | 2478 | path_put(&nd.path); |
2ad94ae6 | 2479 | putname(to); |
1da177e4 | 2480 | out: |
2d8f3038 | 2481 | path_put(&old_path); |
1da177e4 LT |
2482 | |
2483 | return error; | |
2484 | } | |
2485 | ||
3480b257 | 2486 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2487 | { |
c04030e1 | 2488 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2489 | } |
2490 | ||
1da177e4 LT |
2491 | /* |
2492 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2493 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2494 | * Problems: | |
2495 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2496 | * b) race potential - two innocent renames can create a loop together. | |
2497 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2498 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2499 | * story. |
2500 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2501 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2502 | * whether the target exists). Solution: try to be smart with locking |
2503 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2504 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2505 | * move will be locked. Thus we can rank directories by the tree |
2506 | * (ancestors first) and rank all non-directories after them. | |
2507 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2508 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2509 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2510 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2511 | * we'd better make sure that there's no link(2) for them. | |
2512 | * d) some filesystems don't support opened-but-unlinked directories, | |
2513 | * either because of layout or because they are not ready to deal with | |
2514 | * all cases correctly. The latter will be fixed (taking this sort of | |
2515 | * stuff into VFS), but the former is not going away. Solution: the same | |
2516 | * trick as in rmdir(). | |
2517 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2518 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2519 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2520 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2521 | * locking]. |
2522 | */ | |
75c96f85 AB |
2523 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2524 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2525 | { |
2526 | int error = 0; | |
2527 | struct inode *target; | |
2528 | ||
2529 | /* | |
2530 | * If we are going to change the parent - check write permissions, | |
2531 | * we'll need to flip '..'. | |
2532 | */ | |
2533 | if (new_dir != old_dir) { | |
f419a2e3 | 2534 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2535 | if (error) |
2536 | return error; | |
2537 | } | |
2538 | ||
2539 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2540 | if (error) | |
2541 | return error; | |
2542 | ||
2543 | target = new_dentry->d_inode; | |
2544 | if (target) { | |
1b1dcc1b | 2545 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2546 | dentry_unhash(new_dentry); |
2547 | } | |
2548 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2549 | error = -EBUSY; | |
2550 | else | |
2551 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2552 | if (target) { | |
2553 | if (!error) | |
2554 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2555 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2556 | if (d_unhashed(new_dentry)) |
2557 | d_rehash(new_dentry); | |
2558 | dput(new_dentry); | |
2559 | } | |
e31e14ec | 2560 | if (!error) |
349457cc MF |
2561 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2562 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2563 | return error; |
2564 | } | |
2565 | ||
75c96f85 AB |
2566 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2567 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2568 | { |
2569 | struct inode *target; | |
2570 | int error; | |
2571 | ||
2572 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2573 | if (error) | |
2574 | return error; | |
2575 | ||
2576 | dget(new_dentry); | |
2577 | target = new_dentry->d_inode; | |
2578 | if (target) | |
1b1dcc1b | 2579 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2580 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2581 | error = -EBUSY; | |
2582 | else | |
2583 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2584 | if (!error) { | |
349457cc | 2585 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2586 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2587 | } |
2588 | if (target) | |
1b1dcc1b | 2589 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2590 | dput(new_dentry); |
2591 | return error; | |
2592 | } | |
2593 | ||
2594 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2595 | struct inode *new_dir, struct dentry *new_dentry) | |
2596 | { | |
2597 | int error; | |
2598 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2599 | const char *old_name; |
1da177e4 LT |
2600 | |
2601 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2602 | return 0; | |
2603 | ||
2604 | error = may_delete(old_dir, old_dentry, is_dir); | |
2605 | if (error) | |
2606 | return error; | |
2607 | ||
2608 | if (!new_dentry->d_inode) | |
a95164d9 | 2609 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2610 | else |
2611 | error = may_delete(new_dir, new_dentry, is_dir); | |
2612 | if (error) | |
2613 | return error; | |
2614 | ||
acfa4380 | 2615 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2616 | return -EPERM; |
2617 | ||
9e3509e2 JK |
2618 | vfs_dq_init(old_dir); |
2619 | vfs_dq_init(new_dir); | |
1da177e4 | 2620 | |
0eeca283 RL |
2621 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2622 | ||
1da177e4 LT |
2623 | if (is_dir) |
2624 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2625 | else | |
2626 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2627 | if (!error) { | |
0eeca283 | 2628 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2629 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2630 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2631 | } |
0eeca283 RL |
2632 | fsnotify_oldname_free(old_name); |
2633 | ||
1da177e4 LT |
2634 | return error; |
2635 | } | |
2636 | ||
2e4d0924 HC |
2637 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2638 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2639 | { |
2ad94ae6 AV |
2640 | struct dentry *old_dir, *new_dir; |
2641 | struct dentry *old_dentry, *new_dentry; | |
2642 | struct dentry *trap; | |
1da177e4 | 2643 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2644 | char *from; |
2645 | char *to; | |
2646 | int error; | |
1da177e4 | 2647 | |
2ad94ae6 | 2648 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2649 | if (error) |
2650 | goto exit; | |
2651 | ||
2ad94ae6 | 2652 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2653 | if (error) |
2654 | goto exit1; | |
2655 | ||
2656 | error = -EXDEV; | |
4ac91378 | 2657 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2658 | goto exit2; |
2659 | ||
4ac91378 | 2660 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2661 | error = -EBUSY; |
2662 | if (oldnd.last_type != LAST_NORM) | |
2663 | goto exit2; | |
2664 | ||
4ac91378 | 2665 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2666 | if (newnd.last_type != LAST_NORM) |
2667 | goto exit2; | |
2668 | ||
0612d9fb OH |
2669 | oldnd.flags &= ~LOOKUP_PARENT; |
2670 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2671 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2672 | |
1da177e4 LT |
2673 | trap = lock_rename(new_dir, old_dir); |
2674 | ||
49705b77 | 2675 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2676 | error = PTR_ERR(old_dentry); |
2677 | if (IS_ERR(old_dentry)) | |
2678 | goto exit3; | |
2679 | /* source must exist */ | |
2680 | error = -ENOENT; | |
2681 | if (!old_dentry->d_inode) | |
2682 | goto exit4; | |
2683 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2684 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2685 | error = -ENOTDIR; | |
2686 | if (oldnd.last.name[oldnd.last.len]) | |
2687 | goto exit4; | |
2688 | if (newnd.last.name[newnd.last.len]) | |
2689 | goto exit4; | |
2690 | } | |
2691 | /* source should not be ancestor of target */ | |
2692 | error = -EINVAL; | |
2693 | if (old_dentry == trap) | |
2694 | goto exit4; | |
49705b77 | 2695 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2696 | error = PTR_ERR(new_dentry); |
2697 | if (IS_ERR(new_dentry)) | |
2698 | goto exit4; | |
2699 | /* target should not be an ancestor of source */ | |
2700 | error = -ENOTEMPTY; | |
2701 | if (new_dentry == trap) | |
2702 | goto exit5; | |
2703 | ||
9079b1eb DH |
2704 | error = mnt_want_write(oldnd.path.mnt); |
2705 | if (error) | |
2706 | goto exit5; | |
be6d3e56 KT |
2707 | error = security_path_rename(&oldnd.path, old_dentry, |
2708 | &newnd.path, new_dentry); | |
2709 | if (error) | |
2710 | goto exit6; | |
1da177e4 LT |
2711 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2712 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2713 | exit6: |
9079b1eb | 2714 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2715 | exit5: |
2716 | dput(new_dentry); | |
2717 | exit4: | |
2718 | dput(old_dentry); | |
2719 | exit3: | |
2720 | unlock_rename(new_dir, old_dir); | |
2721 | exit2: | |
1d957f9b | 2722 | path_put(&newnd.path); |
2ad94ae6 | 2723 | putname(to); |
1da177e4 | 2724 | exit1: |
1d957f9b | 2725 | path_put(&oldnd.path); |
1da177e4 | 2726 | putname(from); |
2ad94ae6 | 2727 | exit: |
1da177e4 LT |
2728 | return error; |
2729 | } | |
2730 | ||
a26eab24 | 2731 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2732 | { |
2733 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2734 | } | |
2735 | ||
1da177e4 LT |
2736 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2737 | { | |
2738 | int len; | |
2739 | ||
2740 | len = PTR_ERR(link); | |
2741 | if (IS_ERR(link)) | |
2742 | goto out; | |
2743 | ||
2744 | len = strlen(link); | |
2745 | if (len > (unsigned) buflen) | |
2746 | len = buflen; | |
2747 | if (copy_to_user(buffer, link, len)) | |
2748 | len = -EFAULT; | |
2749 | out: | |
2750 | return len; | |
2751 | } | |
2752 | ||
2753 | /* | |
2754 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2755 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2756 | * using) it for any given inode is up to filesystem. | |
2757 | */ | |
2758 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2759 | { | |
2760 | struct nameidata nd; | |
cc314eef | 2761 | void *cookie; |
694a1764 | 2762 | int res; |
cc314eef | 2763 | |
1da177e4 | 2764 | nd.depth = 0; |
cc314eef | 2765 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2766 | if (IS_ERR(cookie)) |
2767 | return PTR_ERR(cookie); | |
2768 | ||
2769 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2770 | if (dentry->d_inode->i_op->put_link) | |
2771 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2772 | return res; | |
1da177e4 LT |
2773 | } |
2774 | ||
2775 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2776 | { | |
2777 | return __vfs_follow_link(nd, link); | |
2778 | } | |
2779 | ||
2780 | /* get the link contents into pagecache */ | |
2781 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2782 | { | |
ebd09abb DG |
2783 | char *kaddr; |
2784 | struct page *page; | |
1da177e4 | 2785 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2786 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2787 | if (IS_ERR(page)) |
6fe6900e | 2788 | return (char*)page; |
1da177e4 | 2789 | *ppage = page; |
ebd09abb DG |
2790 | kaddr = kmap(page); |
2791 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2792 | return kaddr; | |
1da177e4 LT |
2793 | } |
2794 | ||
2795 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2796 | { | |
2797 | struct page *page = NULL; | |
2798 | char *s = page_getlink(dentry, &page); | |
2799 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2800 | if (page) { | |
2801 | kunmap(page); | |
2802 | page_cache_release(page); | |
2803 | } | |
2804 | return res; | |
2805 | } | |
2806 | ||
cc314eef | 2807 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2808 | { |
cc314eef | 2809 | struct page *page = NULL; |
1da177e4 | 2810 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2811 | return page; |
1da177e4 LT |
2812 | } |
2813 | ||
cc314eef | 2814 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2815 | { |
cc314eef LT |
2816 | struct page *page = cookie; |
2817 | ||
2818 | if (page) { | |
1da177e4 LT |
2819 | kunmap(page); |
2820 | page_cache_release(page); | |
1da177e4 LT |
2821 | } |
2822 | } | |
2823 | ||
54566b2c NP |
2824 | /* |
2825 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2826 | */ | |
2827 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2828 | { |
2829 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2830 | struct page *page; |
afddba49 | 2831 | void *fsdata; |
beb497ab | 2832 | int err; |
1da177e4 | 2833 | char *kaddr; |
54566b2c NP |
2834 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2835 | if (nofs) | |
2836 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2837 | |
7e53cac4 | 2838 | retry: |
afddba49 | 2839 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2840 | flags, &page, &fsdata); |
1da177e4 | 2841 | if (err) |
afddba49 NP |
2842 | goto fail; |
2843 | ||
1da177e4 LT |
2844 | kaddr = kmap_atomic(page, KM_USER0); |
2845 | memcpy(kaddr, symname, len-1); | |
2846 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2847 | |
2848 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2849 | page, fsdata); | |
1da177e4 LT |
2850 | if (err < 0) |
2851 | goto fail; | |
afddba49 NP |
2852 | if (err < len-1) |
2853 | goto retry; | |
2854 | ||
1da177e4 LT |
2855 | mark_inode_dirty(inode); |
2856 | return 0; | |
1da177e4 LT |
2857 | fail: |
2858 | return err; | |
2859 | } | |
2860 | ||
0adb25d2 KK |
2861 | int page_symlink(struct inode *inode, const char *symname, int len) |
2862 | { | |
2863 | return __page_symlink(inode, symname, len, | |
54566b2c | 2864 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2865 | } |
2866 | ||
92e1d5be | 2867 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2868 | .readlink = generic_readlink, |
2869 | .follow_link = page_follow_link_light, | |
2870 | .put_link = page_put_link, | |
2871 | }; | |
2872 | ||
2d8f3038 | 2873 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2874 | EXPORT_SYMBOL(follow_down); |
2875 | EXPORT_SYMBOL(follow_up); | |
2876 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2877 | EXPORT_SYMBOL(getname); | |
2878 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2879 | EXPORT_SYMBOL(lookup_one_len); |
2880 | EXPORT_SYMBOL(page_follow_link_light); | |
2881 | EXPORT_SYMBOL(page_put_link); | |
2882 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2883 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2884 | EXPORT_SYMBOL(page_symlink); |
2885 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2886 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2887 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2888 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2889 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2890 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2891 | EXPORT_SYMBOL(unlock_rename); |
2892 | EXPORT_SYMBOL(vfs_create); | |
2893 | EXPORT_SYMBOL(vfs_follow_link); | |
2894 | EXPORT_SYMBOL(vfs_link); | |
2895 | EXPORT_SYMBOL(vfs_mkdir); | |
2896 | EXPORT_SYMBOL(vfs_mknod); | |
2897 | EXPORT_SYMBOL(generic_permission); | |
2898 | EXPORT_SYMBOL(vfs_readlink); | |
2899 | EXPORT_SYMBOL(vfs_rename); | |
2900 | EXPORT_SYMBOL(vfs_rmdir); | |
2901 | EXPORT_SYMBOL(vfs_symlink); | |
2902 | EXPORT_SYMBOL(vfs_unlink); | |
2903 | EXPORT_SYMBOL(dentry_unhash); | |
2904 | EXPORT_SYMBOL(generic_readlink); |