]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/fs.h> | |
21 | #include <linux/namei.h> | |
22 | #include <linux/quotaops.h> | |
23 | #include <linux/pagemap.h> | |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
6146f0d5 | 27 | #include <linux/ima.h> |
1da177e4 LT |
28 | #include <linux/syscalls.h> |
29 | #include <linux/mount.h> | |
30 | #include <linux/audit.h> | |
16f7e0fe | 31 | #include <linux/capability.h> |
834f2a4a | 32 | #include <linux/file.h> |
5590ff0d | 33 | #include <linux/fcntl.h> |
08ce5f16 | 34 | #include <linux/device_cgroup.h> |
5ad4e53b | 35 | #include <linux/fs_struct.h> |
1da177e4 LT |
36 | #include <asm/uaccess.h> |
37 | ||
38 | #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE]) | |
39 | ||
40 | /* [Feb-1997 T. Schoebel-Theuer] | |
41 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
42 | * were necessary because of omirr. The reason is that omirr needs | |
43 | * to know the _real_ pathname, not the user-supplied one, in case | |
44 | * of symlinks (and also when transname replacements occur). | |
45 | * | |
46 | * The new code replaces the old recursive symlink resolution with | |
47 | * an iterative one (in case of non-nested symlink chains). It does | |
48 | * this with calls to <fs>_follow_link(). | |
49 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
50 | * replaced with a single function lookup_dentry() that can handle all | |
51 | * the special cases of the former code. | |
52 | * | |
53 | * With the new dcache, the pathname is stored at each inode, at least as | |
54 | * long as the refcount of the inode is positive. As a side effect, the | |
55 | * size of the dcache depends on the inode cache and thus is dynamic. | |
56 | * | |
57 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
58 | * resolution to correspond with current state of the code. | |
59 | * | |
60 | * Note that the symlink resolution is not *completely* iterative. | |
61 | * There is still a significant amount of tail- and mid- recursion in | |
62 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
63 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
64 | * may return different results than <fs>_follow_link(). Many virtual | |
65 | * filesystems (including /proc) exhibit this behavior. | |
66 | */ | |
67 | ||
68 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
69 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
70 | * and the name already exists in form of a symlink, try to create the new | |
71 | * name indicated by the symlink. The old code always complained that the | |
72 | * name already exists, due to not following the symlink even if its target | |
73 | * is nonexistent. The new semantics affects also mknod() and link() when | |
74 | * the name is a symlink pointing to a non-existant name. | |
75 | * | |
76 | * I don't know which semantics is the right one, since I have no access | |
77 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
78 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
79 | * "old" one. Personally, I think the new semantics is much more logical. | |
80 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
81 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
82 | * and in the old Linux semantics. | |
83 | */ | |
84 | ||
85 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
86 | * semantics. See the comments in "open_namei" and "do_link" below. | |
87 | * | |
88 | * [10-Sep-98 Alan Modra] Another symlink change. | |
89 | */ | |
90 | ||
91 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
92 | * inside the path - always follow. | |
93 | * in the last component in creation/removal/renaming - never follow. | |
94 | * if LOOKUP_FOLLOW passed - follow. | |
95 | * if the pathname has trailing slashes - follow. | |
96 | * otherwise - don't follow. | |
97 | * (applied in that order). | |
98 | * | |
99 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
100 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
101 | * During the 2.4 we need to fix the userland stuff depending on it - | |
102 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
103 | * XEmacs seems to be relying on it... | |
104 | */ | |
105 | /* | |
106 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 107 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
108 | * any extra contention... |
109 | */ | |
110 | ||
111 | /* In order to reduce some races, while at the same time doing additional | |
112 | * checking and hopefully speeding things up, we copy filenames to the | |
113 | * kernel data space before using them.. | |
114 | * | |
115 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
116 | * PATH_MAX includes the nul terminator --RR. | |
117 | */ | |
858119e1 | 118 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
119 | { |
120 | int retval; | |
121 | unsigned long len = PATH_MAX; | |
122 | ||
123 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
124 | if ((unsigned long) filename >= TASK_SIZE) | |
125 | return -EFAULT; | |
126 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
127 | len = TASK_SIZE - (unsigned long) filename; | |
128 | } | |
129 | ||
130 | retval = strncpy_from_user(page, filename, len); | |
131 | if (retval > 0) { | |
132 | if (retval < len) | |
133 | return 0; | |
134 | return -ENAMETOOLONG; | |
135 | } else if (!retval) | |
136 | retval = -ENOENT; | |
137 | return retval; | |
138 | } | |
139 | ||
140 | char * getname(const char __user * filename) | |
141 | { | |
142 | char *tmp, *result; | |
143 | ||
144 | result = ERR_PTR(-ENOMEM); | |
145 | tmp = __getname(); | |
146 | if (tmp) { | |
147 | int retval = do_getname(filename, tmp); | |
148 | ||
149 | result = tmp; | |
150 | if (retval < 0) { | |
151 | __putname(tmp); | |
152 | result = ERR_PTR(retval); | |
153 | } | |
154 | } | |
155 | audit_getname(result); | |
156 | return result; | |
157 | } | |
158 | ||
159 | #ifdef CONFIG_AUDITSYSCALL | |
160 | void putname(const char *name) | |
161 | { | |
5ac3a9c2 | 162 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
163 | audit_putname(name); |
164 | else | |
165 | __putname(name); | |
166 | } | |
167 | EXPORT_SYMBOL(putname); | |
168 | #endif | |
169 | ||
5909ccaa LT |
170 | /* |
171 | * This does basic POSIX ACL permission checking | |
1da177e4 | 172 | */ |
5909ccaa | 173 | static int acl_permission_check(struct inode *inode, int mask, |
1da177e4 LT |
174 | int (*check_acl)(struct inode *inode, int mask)) |
175 | { | |
176 | umode_t mode = inode->i_mode; | |
177 | ||
e6305c43 AV |
178 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
179 | ||
da9592ed | 180 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
181 | mode >>= 6; |
182 | else { | |
183 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
184 | int error = check_acl(inode, mask); | |
5909ccaa | 185 | if (error != -EAGAIN) |
1da177e4 LT |
186 | return error; |
187 | } | |
188 | ||
189 | if (in_group_p(inode->i_gid)) | |
190 | mode >>= 3; | |
191 | } | |
192 | ||
193 | /* | |
194 | * If the DACs are ok we don't need any capability check. | |
195 | */ | |
e6305c43 | 196 | if ((mask & ~mode) == 0) |
1da177e4 | 197 | return 0; |
5909ccaa LT |
198 | return -EACCES; |
199 | } | |
200 | ||
201 | /** | |
202 | * generic_permission - check for access rights on a Posix-like filesystem | |
203 | * @inode: inode to check access rights for | |
204 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
205 | * @check_acl: optional callback to check for Posix ACLs | |
206 | * | |
207 | * Used to check for read/write/execute permissions on a file. | |
208 | * We use "fsuid" for this, letting us set arbitrary permissions | |
209 | * for filesystem access without changing the "normal" uids which | |
210 | * are used for other things.. | |
211 | */ | |
212 | int generic_permission(struct inode *inode, int mask, | |
213 | int (*check_acl)(struct inode *inode, int mask)) | |
214 | { | |
215 | int ret; | |
216 | ||
217 | /* | |
218 | * Do the basic POSIX ACL permission checks. | |
219 | */ | |
220 | ret = acl_permission_check(inode, mask, check_acl); | |
221 | if (ret != -EACCES) | |
222 | return ret; | |
1da177e4 | 223 | |
1da177e4 LT |
224 | /* |
225 | * Read/write DACs are always overridable. | |
226 | * Executable DACs are overridable if at least one exec bit is set. | |
227 | */ | |
f696a365 | 228 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
229 | if (capable(CAP_DAC_OVERRIDE)) |
230 | return 0; | |
231 | ||
232 | /* | |
233 | * Searching includes executable on directories, else just read. | |
234 | */ | |
235 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) | |
236 | if (capable(CAP_DAC_READ_SEARCH)) | |
237 | return 0; | |
238 | ||
239 | return -EACCES; | |
240 | } | |
241 | ||
cb23beb5 CH |
242 | /** |
243 | * inode_permission - check for access rights to a given inode | |
244 | * @inode: inode to check permission on | |
245 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
246 | * | |
247 | * Used to check for read/write/execute permissions on an inode. | |
248 | * We use "fsuid" for this, letting us set arbitrary permissions | |
249 | * for filesystem access without changing the "normal" uids which | |
250 | * are used for other things. | |
251 | */ | |
f419a2e3 | 252 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 253 | { |
e6305c43 | 254 | int retval; |
1da177e4 LT |
255 | |
256 | if (mask & MAY_WRITE) { | |
22590e41 | 257 | umode_t mode = inode->i_mode; |
1da177e4 LT |
258 | |
259 | /* | |
260 | * Nobody gets write access to a read-only fs. | |
261 | */ | |
262 | if (IS_RDONLY(inode) && | |
263 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
264 | return -EROFS; | |
265 | ||
266 | /* | |
267 | * Nobody gets write access to an immutable file. | |
268 | */ | |
269 | if (IS_IMMUTABLE(inode)) | |
270 | return -EACCES; | |
271 | } | |
272 | ||
acfa4380 | 273 | if (inode->i_op->permission) |
b77b0646 | 274 | retval = inode->i_op->permission(inode, mask); |
f696a365 | 275 | else |
5909ccaa | 276 | retval = generic_permission(inode, mask, inode->i_op->check_acl); |
f696a365 | 277 | |
1da177e4 LT |
278 | if (retval) |
279 | return retval; | |
280 | ||
08ce5f16 SH |
281 | retval = devcgroup_inode_permission(inode, mask); |
282 | if (retval) | |
283 | return retval; | |
284 | ||
e6305c43 | 285 | return security_inode_permission(inode, |
f418b006 | 286 | mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND)); |
1da177e4 LT |
287 | } |
288 | ||
8c744fb8 CH |
289 | /** |
290 | * file_permission - check for additional access rights to a given file | |
291 | * @file: file to check access rights for | |
292 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
293 | * | |
294 | * Used to check for read/write/execute permissions on an already opened | |
295 | * file. | |
296 | * | |
297 | * Note: | |
298 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 299 | * be done using inode_permission(). |
8c744fb8 CH |
300 | */ |
301 | int file_permission(struct file *file, int mask) | |
302 | { | |
f419a2e3 | 303 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
304 | } |
305 | ||
1da177e4 LT |
306 | /* |
307 | * get_write_access() gets write permission for a file. | |
308 | * put_write_access() releases this write permission. | |
309 | * This is used for regular files. | |
310 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
311 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
312 | * can have the following values: | |
313 | * 0: no writers, no VM_DENYWRITE mappings | |
314 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
315 | * > 0: (i_writecount) users are writing to the file. | |
316 | * | |
317 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
318 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
319 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
320 | * to do the change if sign is wrong. Exclusion between them is provided by | |
321 | * the inode->i_lock spinlock. | |
322 | */ | |
323 | ||
324 | int get_write_access(struct inode * inode) | |
325 | { | |
326 | spin_lock(&inode->i_lock); | |
327 | if (atomic_read(&inode->i_writecount) < 0) { | |
328 | spin_unlock(&inode->i_lock); | |
329 | return -ETXTBSY; | |
330 | } | |
331 | atomic_inc(&inode->i_writecount); | |
332 | spin_unlock(&inode->i_lock); | |
333 | ||
334 | return 0; | |
335 | } | |
336 | ||
337 | int deny_write_access(struct file * file) | |
338 | { | |
0f7fc9e4 | 339 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
340 | |
341 | spin_lock(&inode->i_lock); | |
342 | if (atomic_read(&inode->i_writecount) > 0) { | |
343 | spin_unlock(&inode->i_lock); | |
344 | return -ETXTBSY; | |
345 | } | |
346 | atomic_dec(&inode->i_writecount); | |
347 | spin_unlock(&inode->i_lock); | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
5dd784d0 JB |
352 | /** |
353 | * path_get - get a reference to a path | |
354 | * @path: path to get the reference to | |
355 | * | |
356 | * Given a path increment the reference count to the dentry and the vfsmount. | |
357 | */ | |
358 | void path_get(struct path *path) | |
359 | { | |
360 | mntget(path->mnt); | |
361 | dget(path->dentry); | |
362 | } | |
363 | EXPORT_SYMBOL(path_get); | |
364 | ||
1d957f9b JB |
365 | /** |
366 | * path_put - put a reference to a path | |
367 | * @path: path to put the reference to | |
368 | * | |
369 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
370 | */ | |
371 | void path_put(struct path *path) | |
1da177e4 | 372 | { |
1d957f9b JB |
373 | dput(path->dentry); |
374 | mntput(path->mnt); | |
1da177e4 | 375 | } |
1d957f9b | 376 | EXPORT_SYMBOL(path_put); |
1da177e4 | 377 | |
834f2a4a TM |
378 | /** |
379 | * release_open_intent - free up open intent resources | |
380 | * @nd: pointer to nameidata | |
381 | */ | |
382 | void release_open_intent(struct nameidata *nd) | |
383 | { | |
0f7fc9e4 | 384 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
385 | put_filp(nd->intent.open.file); |
386 | else | |
387 | fput(nd->intent.open.file); | |
388 | } | |
389 | ||
bcdc5e01 IK |
390 | static inline struct dentry * |
391 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
392 | { | |
393 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
394 | if (unlikely(status <= 0)) { | |
395 | /* | |
396 | * The dentry failed validation. | |
397 | * If d_revalidate returned 0 attempt to invalidate | |
398 | * the dentry otherwise d_revalidate is asking us | |
399 | * to return a fail status. | |
400 | */ | |
401 | if (!status) { | |
402 | if (!d_invalidate(dentry)) { | |
403 | dput(dentry); | |
404 | dentry = NULL; | |
405 | } | |
406 | } else { | |
407 | dput(dentry); | |
408 | dentry = ERR_PTR(status); | |
409 | } | |
410 | } | |
411 | return dentry; | |
412 | } | |
413 | ||
1da177e4 LT |
414 | /* |
415 | * Short-cut version of permission(), for calling by | |
416 | * path_walk(), when dcache lock is held. Combines parts | |
417 | * of permission() and generic_permission(), and tests ONLY for | |
418 | * MAY_EXEC permission. | |
419 | * | |
420 | * If appropriate, check DAC only. If not appropriate, or | |
421 | * short-cut DAC fails, then call permission() to do more | |
422 | * complete permission check. | |
423 | */ | |
672b16b2 | 424 | static int exec_permission_lite(struct inode *inode) |
1da177e4 | 425 | { |
5909ccaa | 426 | int ret; |
1da177e4 | 427 | |
cb9179ea | 428 | if (inode->i_op->permission) { |
5909ccaa | 429 | ret = inode->i_op->permission(inode, MAY_EXEC); |
cb9179ea LT |
430 | if (!ret) |
431 | goto ok; | |
432 | return ret; | |
433 | } | |
5909ccaa LT |
434 | ret = acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl); |
435 | if (!ret) | |
1da177e4 LT |
436 | goto ok; |
437 | ||
f1ac9f6b | 438 | if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
439 | goto ok; |
440 | ||
5909ccaa | 441 | return ret; |
1da177e4 | 442 | ok: |
b77b0646 | 443 | return security_inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
444 | } |
445 | ||
2a737871 AV |
446 | static __always_inline void set_root(struct nameidata *nd) |
447 | { | |
448 | if (!nd->root.mnt) { | |
449 | struct fs_struct *fs = current->fs; | |
450 | read_lock(&fs->lock); | |
451 | nd->root = fs->root; | |
452 | path_get(&nd->root); | |
453 | read_unlock(&fs->lock); | |
454 | } | |
455 | } | |
456 | ||
6de88d72 AV |
457 | static int link_path_walk(const char *, struct nameidata *); |
458 | ||
f1662356 | 459 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 LT |
460 | { |
461 | int res = 0; | |
462 | char *name; | |
463 | if (IS_ERR(link)) | |
464 | goto fail; | |
465 | ||
466 | if (*link == '/') { | |
2a737871 | 467 | set_root(nd); |
1d957f9b | 468 | path_put(&nd->path); |
2a737871 AV |
469 | nd->path = nd->root; |
470 | path_get(&nd->root); | |
1da177e4 | 471 | } |
b4091d5f | 472 | |
1da177e4 | 473 | res = link_path_walk(link, nd); |
1da177e4 LT |
474 | if (nd->depth || res || nd->last_type!=LAST_NORM) |
475 | return res; | |
476 | /* | |
477 | * If it is an iterative symlinks resolution in open_namei() we | |
478 | * have to copy the last component. And all that crap because of | |
479 | * bloody create() on broken symlinks. Furrfu... | |
480 | */ | |
481 | name = __getname(); | |
482 | if (unlikely(!name)) { | |
1d957f9b | 483 | path_put(&nd->path); |
1da177e4 LT |
484 | return -ENOMEM; |
485 | } | |
486 | strcpy(name, nd->last.name); | |
487 | nd->last.name = name; | |
488 | return 0; | |
489 | fail: | |
1d957f9b | 490 | path_put(&nd->path); |
1da177e4 LT |
491 | return PTR_ERR(link); |
492 | } | |
493 | ||
1d957f9b | 494 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
495 | { |
496 | dput(path->dentry); | |
4ac91378 | 497 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
498 | mntput(path->mnt); |
499 | } | |
500 | ||
501 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
502 | { | |
4ac91378 JB |
503 | dput(nd->path.dentry); |
504 | if (nd->path.mnt != path->mnt) | |
505 | mntput(nd->path.mnt); | |
506 | nd->path.mnt = path->mnt; | |
507 | nd->path.dentry = path->dentry; | |
051d3812 IK |
508 | } |
509 | ||
f1662356 | 510 | static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
511 | { |
512 | int error; | |
cc314eef | 513 | void *cookie; |
cd4e91d3 | 514 | struct dentry *dentry = path->dentry; |
1da177e4 | 515 | |
d671a1cb | 516 | touch_atime(path->mnt, dentry); |
1da177e4 | 517 | nd_set_link(nd, NULL); |
cd4e91d3 | 518 | |
4ac91378 | 519 | if (path->mnt != nd->path.mnt) { |
051d3812 IK |
520 | path_to_nameidata(path, nd); |
521 | dget(dentry); | |
522 | } | |
523 | mntget(path->mnt); | |
cc314eef LT |
524 | cookie = dentry->d_inode->i_op->follow_link(dentry, nd); |
525 | error = PTR_ERR(cookie); | |
526 | if (!IS_ERR(cookie)) { | |
1da177e4 | 527 | char *s = nd_get_link(nd); |
cc314eef | 528 | error = 0; |
1da177e4 LT |
529 | if (s) |
530 | error = __vfs_follow_link(nd, s); | |
531 | if (dentry->d_inode->i_op->put_link) | |
cc314eef | 532 | dentry->d_inode->i_op->put_link(dentry, nd, cookie); |
1da177e4 | 533 | } |
1da177e4 LT |
534 | return error; |
535 | } | |
536 | ||
537 | /* | |
538 | * This limits recursive symlink follows to 8, while | |
539 | * limiting consecutive symlinks to 40. | |
540 | * | |
541 | * Without that kind of total limit, nasty chains of consecutive | |
542 | * symlinks can cause almost arbitrarily long lookups. | |
543 | */ | |
90ebe565 | 544 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 LT |
545 | { |
546 | int err = -ELOOP; | |
547 | if (current->link_count >= MAX_NESTED_LINKS) | |
548 | goto loop; | |
549 | if (current->total_link_count >= 40) | |
550 | goto loop; | |
551 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
552 | cond_resched(); | |
90ebe565 | 553 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
554 | if (err) |
555 | goto loop; | |
556 | current->link_count++; | |
557 | current->total_link_count++; | |
558 | nd->depth++; | |
cd4e91d3 | 559 | err = __do_follow_link(path, nd); |
258fa999 | 560 | path_put(path); |
839d9f93 AV |
561 | current->link_count--; |
562 | nd->depth--; | |
1da177e4 LT |
563 | return err; |
564 | loop: | |
1d957f9b JB |
565 | path_put_conditional(path, nd); |
566 | path_put(&nd->path); | |
1da177e4 LT |
567 | return err; |
568 | } | |
569 | ||
bab77ebf | 570 | int follow_up(struct path *path) |
1da177e4 LT |
571 | { |
572 | struct vfsmount *parent; | |
573 | struct dentry *mountpoint; | |
574 | spin_lock(&vfsmount_lock); | |
bab77ebf AV |
575 | parent = path->mnt->mnt_parent; |
576 | if (parent == path->mnt) { | |
1da177e4 LT |
577 | spin_unlock(&vfsmount_lock); |
578 | return 0; | |
579 | } | |
580 | mntget(parent); | |
bab77ebf | 581 | mountpoint = dget(path->mnt->mnt_mountpoint); |
1da177e4 | 582 | spin_unlock(&vfsmount_lock); |
bab77ebf AV |
583 | dput(path->dentry); |
584 | path->dentry = mountpoint; | |
585 | mntput(path->mnt); | |
586 | path->mnt = parent; | |
1da177e4 LT |
587 | return 1; |
588 | } | |
589 | ||
590 | /* no need for dcache_lock, as serialization is taken care in | |
591 | * namespace.c | |
592 | */ | |
463ffb2e AV |
593 | static int __follow_mount(struct path *path) |
594 | { | |
595 | int res = 0; | |
596 | while (d_mountpoint(path->dentry)) { | |
1c755af4 | 597 | struct vfsmount *mounted = lookup_mnt(path); |
463ffb2e AV |
598 | if (!mounted) |
599 | break; | |
600 | dput(path->dentry); | |
601 | if (res) | |
602 | mntput(path->mnt); | |
603 | path->mnt = mounted; | |
604 | path->dentry = dget(mounted->mnt_root); | |
605 | res = 1; | |
606 | } | |
607 | return res; | |
608 | } | |
609 | ||
79ed0226 | 610 | static void follow_mount(struct path *path) |
1da177e4 | 611 | { |
79ed0226 | 612 | while (d_mountpoint(path->dentry)) { |
1c755af4 | 613 | struct vfsmount *mounted = lookup_mnt(path); |
1da177e4 LT |
614 | if (!mounted) |
615 | break; | |
79ed0226 AV |
616 | dput(path->dentry); |
617 | mntput(path->mnt); | |
618 | path->mnt = mounted; | |
619 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 | 620 | } |
1da177e4 LT |
621 | } |
622 | ||
623 | /* no need for dcache_lock, as serialization is taken care in | |
624 | * namespace.c | |
625 | */ | |
9393bd07 | 626 | int follow_down(struct path *path) |
1da177e4 LT |
627 | { |
628 | struct vfsmount *mounted; | |
629 | ||
1c755af4 | 630 | mounted = lookup_mnt(path); |
1da177e4 | 631 | if (mounted) { |
9393bd07 AV |
632 | dput(path->dentry); |
633 | mntput(path->mnt); | |
634 | path->mnt = mounted; | |
635 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
636 | return 1; |
637 | } | |
638 | return 0; | |
639 | } | |
640 | ||
f1662356 | 641 | static __always_inline void follow_dotdot(struct nameidata *nd) |
1da177e4 | 642 | { |
2a737871 | 643 | set_root(nd); |
e518ddb7 | 644 | |
1da177e4 LT |
645 | while(1) { |
646 | struct vfsmount *parent; | |
4ac91378 | 647 | struct dentry *old = nd->path.dentry; |
1da177e4 | 648 | |
2a737871 AV |
649 | if (nd->path.dentry == nd->root.dentry && |
650 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
651 | break; |
652 | } | |
1da177e4 | 653 | spin_lock(&dcache_lock); |
4ac91378 JB |
654 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
655 | nd->path.dentry = dget(nd->path.dentry->d_parent); | |
1da177e4 LT |
656 | spin_unlock(&dcache_lock); |
657 | dput(old); | |
658 | break; | |
659 | } | |
660 | spin_unlock(&dcache_lock); | |
661 | spin_lock(&vfsmount_lock); | |
4ac91378 JB |
662 | parent = nd->path.mnt->mnt_parent; |
663 | if (parent == nd->path.mnt) { | |
1da177e4 LT |
664 | spin_unlock(&vfsmount_lock); |
665 | break; | |
666 | } | |
667 | mntget(parent); | |
4ac91378 | 668 | nd->path.dentry = dget(nd->path.mnt->mnt_mountpoint); |
1da177e4 LT |
669 | spin_unlock(&vfsmount_lock); |
670 | dput(old); | |
4ac91378 JB |
671 | mntput(nd->path.mnt); |
672 | nd->path.mnt = parent; | |
1da177e4 | 673 | } |
79ed0226 | 674 | follow_mount(&nd->path); |
1da177e4 LT |
675 | } |
676 | ||
1da177e4 LT |
677 | /* |
678 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
679 | * small and for now I'd prefer to have fast path as straight as possible. | |
680 | * It _is_ time-critical. | |
681 | */ | |
682 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
683 | struct path *path) | |
684 | { | |
4ac91378 | 685 | struct vfsmount *mnt = nd->path.mnt; |
6e6b1bd1 AV |
686 | struct dentry *dentry, *parent; |
687 | struct inode *dir; | |
3cac260a AV |
688 | /* |
689 | * See if the low-level filesystem might want | |
690 | * to use its own hash.. | |
691 | */ | |
692 | if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { | |
693 | int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, name); | |
694 | if (err < 0) | |
695 | return err; | |
696 | } | |
1da177e4 | 697 | |
3cac260a | 698 | dentry = __d_lookup(nd->path.dentry, name); |
1da177e4 LT |
699 | if (!dentry) |
700 | goto need_lookup; | |
701 | if (dentry->d_op && dentry->d_op->d_revalidate) | |
702 | goto need_revalidate; | |
703 | done: | |
704 | path->mnt = mnt; | |
705 | path->dentry = dentry; | |
634ee701 | 706 | __follow_mount(path); |
1da177e4 LT |
707 | return 0; |
708 | ||
709 | need_lookup: | |
6e6b1bd1 AV |
710 | parent = nd->path.dentry; |
711 | dir = parent->d_inode; | |
712 | ||
713 | mutex_lock(&dir->i_mutex); | |
714 | /* | |
715 | * First re-do the cached lookup just in case it was created | |
716 | * while we waited for the directory semaphore.. | |
717 | * | |
718 | * FIXME! This could use version numbering or similar to | |
719 | * avoid unnecessary cache lookups. | |
720 | * | |
721 | * The "dcache_lock" is purely to protect the RCU list walker | |
722 | * from concurrent renames at this point (we mustn't get false | |
723 | * negatives from the RCU list walk here, unlike the optimistic | |
724 | * fast walk). | |
725 | * | |
726 | * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup | |
727 | */ | |
728 | dentry = d_lookup(parent, name); | |
729 | if (!dentry) { | |
730 | struct dentry *new; | |
731 | ||
732 | /* Don't create child dentry for a dead directory. */ | |
733 | dentry = ERR_PTR(-ENOENT); | |
734 | if (IS_DEADDIR(dir)) | |
735 | goto out_unlock; | |
736 | ||
737 | new = d_alloc(parent, name); | |
738 | dentry = ERR_PTR(-ENOMEM); | |
739 | if (new) { | |
740 | dentry = dir->i_op->lookup(dir, new, nd); | |
741 | if (dentry) | |
742 | dput(new); | |
743 | else | |
744 | dentry = new; | |
745 | } | |
746 | out_unlock: | |
747 | mutex_unlock(&dir->i_mutex); | |
748 | if (IS_ERR(dentry)) | |
749 | goto fail; | |
750 | goto done; | |
751 | } | |
752 | ||
753 | /* | |
754 | * Uhhuh! Nasty case: the cache was re-populated while | |
755 | * we waited on the semaphore. Need to revalidate. | |
756 | */ | |
757 | mutex_unlock(&dir->i_mutex); | |
758 | if (dentry->d_op && dentry->d_op->d_revalidate) { | |
759 | dentry = do_revalidate(dentry, nd); | |
760 | if (!dentry) | |
761 | dentry = ERR_PTR(-ENOENT); | |
762 | } | |
1da177e4 LT |
763 | if (IS_ERR(dentry)) |
764 | goto fail; | |
765 | goto done; | |
766 | ||
767 | need_revalidate: | |
bcdc5e01 IK |
768 | dentry = do_revalidate(dentry, nd); |
769 | if (!dentry) | |
770 | goto need_lookup; | |
771 | if (IS_ERR(dentry)) | |
772 | goto fail; | |
773 | goto done; | |
1da177e4 LT |
774 | |
775 | fail: | |
776 | return PTR_ERR(dentry); | |
777 | } | |
778 | ||
779 | /* | |
780 | * Name resolution. | |
ea3834d9 PM |
781 | * This is the basic name resolution function, turning a pathname into |
782 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 783 | * |
ea3834d9 PM |
784 | * Returns 0 and nd will have valid dentry and mnt on success. |
785 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 786 | */ |
6de88d72 | 787 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
788 | { |
789 | struct path next; | |
790 | struct inode *inode; | |
791 | int err; | |
792 | unsigned int lookup_flags = nd->flags; | |
793 | ||
794 | while (*name=='/') | |
795 | name++; | |
796 | if (!*name) | |
797 | goto return_reval; | |
798 | ||
4ac91378 | 799 | inode = nd->path.dentry->d_inode; |
1da177e4 | 800 | if (nd->depth) |
f55eab82 | 801 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
802 | |
803 | /* At this point we know we have a real path component. */ | |
804 | for(;;) { | |
805 | unsigned long hash; | |
806 | struct qstr this; | |
807 | unsigned int c; | |
808 | ||
cdce5d6b | 809 | nd->flags |= LOOKUP_CONTINUE; |
672b16b2 | 810 | err = exec_permission_lite(inode); |
1da177e4 LT |
811 | if (err) |
812 | break; | |
813 | ||
814 | this.name = name; | |
815 | c = *(const unsigned char *)name; | |
816 | ||
817 | hash = init_name_hash(); | |
818 | do { | |
819 | name++; | |
820 | hash = partial_name_hash(c, hash); | |
821 | c = *(const unsigned char *)name; | |
822 | } while (c && (c != '/')); | |
823 | this.len = name - (const char *) this.name; | |
824 | this.hash = end_name_hash(hash); | |
825 | ||
826 | /* remove trailing slashes? */ | |
827 | if (!c) | |
828 | goto last_component; | |
829 | while (*++name == '/'); | |
830 | if (!*name) | |
831 | goto last_with_slashes; | |
832 | ||
833 | /* | |
834 | * "." and ".." are special - ".." especially so because it has | |
835 | * to be able to know about the current root directory and | |
836 | * parent relationships. | |
837 | */ | |
838 | if (this.name[0] == '.') switch (this.len) { | |
839 | default: | |
840 | break; | |
841 | case 2: | |
842 | if (this.name[1] != '.') | |
843 | break; | |
58c465eb | 844 | follow_dotdot(nd); |
4ac91378 | 845 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
846 | /* fallthrough */ |
847 | case 1: | |
848 | continue; | |
849 | } | |
1da177e4 LT |
850 | /* This does the actual lookups.. */ |
851 | err = do_lookup(nd, &this, &next); | |
852 | if (err) | |
853 | break; | |
1da177e4 LT |
854 | |
855 | err = -ENOENT; | |
856 | inode = next.dentry->d_inode; | |
857 | if (!inode) | |
858 | goto out_dput; | |
1da177e4 LT |
859 | |
860 | if (inode->i_op->follow_link) { | |
90ebe565 | 861 | err = do_follow_link(&next, nd); |
1da177e4 LT |
862 | if (err) |
863 | goto return_err; | |
864 | err = -ENOENT; | |
4ac91378 | 865 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
866 | if (!inode) |
867 | break; | |
09dd17d3 MS |
868 | } else |
869 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
870 | err = -ENOTDIR; |
871 | if (!inode->i_op->lookup) | |
872 | break; | |
873 | continue; | |
874 | /* here ends the main loop */ | |
875 | ||
876 | last_with_slashes: | |
877 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
878 | last_component: | |
f55eab82 TM |
879 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
880 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
881 | if (lookup_flags & LOOKUP_PARENT) |
882 | goto lookup_parent; | |
883 | if (this.name[0] == '.') switch (this.len) { | |
884 | default: | |
885 | break; | |
886 | case 2: | |
887 | if (this.name[1] != '.') | |
888 | break; | |
58c465eb | 889 | follow_dotdot(nd); |
4ac91378 | 890 | inode = nd->path.dentry->d_inode; |
1da177e4 LT |
891 | /* fallthrough */ |
892 | case 1: | |
893 | goto return_reval; | |
894 | } | |
1da177e4 LT |
895 | err = do_lookup(nd, &this, &next); |
896 | if (err) | |
897 | break; | |
1da177e4 LT |
898 | inode = next.dentry->d_inode; |
899 | if ((lookup_flags & LOOKUP_FOLLOW) | |
acfa4380 | 900 | && inode && inode->i_op->follow_link) { |
90ebe565 | 901 | err = do_follow_link(&next, nd); |
1da177e4 LT |
902 | if (err) |
903 | goto return_err; | |
4ac91378 | 904 | inode = nd->path.dentry->d_inode; |
09dd17d3 MS |
905 | } else |
906 | path_to_nameidata(&next, nd); | |
1da177e4 LT |
907 | err = -ENOENT; |
908 | if (!inode) | |
909 | break; | |
910 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
911 | err = -ENOTDIR; | |
acfa4380 | 912 | if (!inode->i_op->lookup) |
1da177e4 LT |
913 | break; |
914 | } | |
915 | goto return_base; | |
916 | lookup_parent: | |
917 | nd->last = this; | |
918 | nd->last_type = LAST_NORM; | |
919 | if (this.name[0] != '.') | |
920 | goto return_base; | |
921 | if (this.len == 1) | |
922 | nd->last_type = LAST_DOT; | |
923 | else if (this.len == 2 && this.name[1] == '.') | |
924 | nd->last_type = LAST_DOTDOT; | |
925 | else | |
926 | goto return_base; | |
927 | return_reval: | |
928 | /* | |
929 | * We bypassed the ordinary revalidation routines. | |
930 | * We may need to check the cached dentry for staleness. | |
931 | */ | |
4ac91378 JB |
932 | if (nd->path.dentry && nd->path.dentry->d_sb && |
933 | (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { | |
1da177e4 LT |
934 | err = -ESTALE; |
935 | /* Note: we do not d_invalidate() */ | |
4ac91378 JB |
936 | if (!nd->path.dentry->d_op->d_revalidate( |
937 | nd->path.dentry, nd)) | |
1da177e4 LT |
938 | break; |
939 | } | |
940 | return_base: | |
941 | return 0; | |
942 | out_dput: | |
1d957f9b | 943 | path_put_conditional(&next, nd); |
1da177e4 LT |
944 | break; |
945 | } | |
1d957f9b | 946 | path_put(&nd->path); |
1da177e4 LT |
947 | return_err: |
948 | return err; | |
949 | } | |
950 | ||
fc9b52cd | 951 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 952 | { |
6de88d72 AV |
953 | struct path save = nd->path; |
954 | int result; | |
955 | ||
1da177e4 | 956 | current->total_link_count = 0; |
6de88d72 AV |
957 | |
958 | /* make sure the stuff we saved doesn't go away */ | |
959 | path_get(&save); | |
960 | ||
961 | result = link_path_walk(name, nd); | |
962 | if (result == -ESTALE) { | |
963 | /* nd->path had been dropped */ | |
964 | current->total_link_count = 0; | |
965 | nd->path = save; | |
966 | path_get(&nd->path); | |
967 | nd->flags |= LOOKUP_REVAL; | |
968 | result = link_path_walk(name, nd); | |
969 | } | |
970 | ||
971 | path_put(&save); | |
972 | ||
973 | return result; | |
1da177e4 LT |
974 | } |
975 | ||
9b4a9b14 | 976 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 977 | { |
ea3834d9 | 978 | int retval = 0; |
170aa3d0 UD |
979 | int fput_needed; |
980 | struct file *file; | |
1da177e4 LT |
981 | |
982 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
983 | nd->flags = flags; | |
984 | nd->depth = 0; | |
2a737871 | 985 | nd->root.mnt = NULL; |
1da177e4 | 986 | |
1da177e4 | 987 | if (*name=='/') { |
2a737871 AV |
988 | set_root(nd); |
989 | nd->path = nd->root; | |
990 | path_get(&nd->root); | |
5590ff0d | 991 | } else if (dfd == AT_FDCWD) { |
2a737871 | 992 | struct fs_struct *fs = current->fs; |
e518ddb7 | 993 | read_lock(&fs->lock); |
6ac08c39 JB |
994 | nd->path = fs->pwd; |
995 | path_get(&fs->pwd); | |
e518ddb7 | 996 | read_unlock(&fs->lock); |
5590ff0d | 997 | } else { |
5590ff0d UD |
998 | struct dentry *dentry; |
999 | ||
1000 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1001 | retval = -EBADF; |
1002 | if (!file) | |
6d09bb62 | 1003 | goto out_fail; |
5590ff0d | 1004 | |
0f7fc9e4 | 1005 | dentry = file->f_path.dentry; |
5590ff0d | 1006 | |
170aa3d0 UD |
1007 | retval = -ENOTDIR; |
1008 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1009 | goto fput_fail; |
5590ff0d UD |
1010 | |
1011 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1012 | if (retval) |
6d09bb62 | 1013 | goto fput_fail; |
5590ff0d | 1014 | |
5dd784d0 JB |
1015 | nd->path = file->f_path; |
1016 | path_get(&file->f_path); | |
5590ff0d UD |
1017 | |
1018 | fput_light(file, fput_needed); | |
1da177e4 | 1019 | } |
9b4a9b14 | 1020 | return 0; |
2dfdd266 | 1021 | |
9b4a9b14 AV |
1022 | fput_fail: |
1023 | fput_light(file, fput_needed); | |
1024 | out_fail: | |
1025 | return retval; | |
1026 | } | |
1027 | ||
1028 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1029 | static int do_path_lookup(int dfd, const char *name, | |
1030 | unsigned int flags, struct nameidata *nd) | |
1031 | { | |
1032 | int retval = path_init(dfd, name, flags, nd); | |
1033 | if (!retval) | |
1034 | retval = path_walk(name, nd); | |
4ac91378 JB |
1035 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1036 | nd->path.dentry->d_inode)) | |
1037 | audit_inode(name, nd->path.dentry); | |
2a737871 AV |
1038 | if (nd->root.mnt) { |
1039 | path_put(&nd->root); | |
1040 | nd->root.mnt = NULL; | |
1041 | } | |
170aa3d0 | 1042 | return retval; |
1da177e4 LT |
1043 | } |
1044 | ||
fc9b52cd | 1045 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1046 | struct nameidata *nd) |
1047 | { | |
1048 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1049 | } | |
1050 | ||
d1811465 AV |
1051 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1052 | { | |
1053 | struct nameidata nd; | |
1054 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1055 | if (!res) | |
1056 | *path = nd.path; | |
1057 | return res; | |
1058 | } | |
1059 | ||
16f18200 JJS |
1060 | /** |
1061 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1062 | * @dentry: pointer to dentry of the base directory | |
1063 | * @mnt: pointer to vfs mount of the base directory | |
1064 | * @name: pointer to file name | |
1065 | * @flags: lookup flags | |
1066 | * @nd: pointer to nameidata | |
1067 | */ | |
1068 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1069 | const char *name, unsigned int flags, | |
1070 | struct nameidata *nd) | |
1071 | { | |
1072 | int retval; | |
1073 | ||
1074 | /* same as do_path_lookup */ | |
1075 | nd->last_type = LAST_ROOT; | |
1076 | nd->flags = flags; | |
1077 | nd->depth = 0; | |
1078 | ||
c8e7f449 JB |
1079 | nd->path.dentry = dentry; |
1080 | nd->path.mnt = mnt; | |
1081 | path_get(&nd->path); | |
5b857119 AV |
1082 | nd->root = nd->path; |
1083 | path_get(&nd->root); | |
16f18200 JJS |
1084 | |
1085 | retval = path_walk(name, nd); | |
4ac91378 JB |
1086 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
1087 | nd->path.dentry->d_inode)) | |
1088 | audit_inode(name, nd->path.dentry); | |
16f18200 | 1089 | |
5b857119 AV |
1090 | path_put(&nd->root); |
1091 | nd->root.mnt = NULL; | |
16f18200 | 1092 | |
2a737871 | 1093 | return retval; |
16f18200 JJS |
1094 | } |
1095 | ||
eead1911 CH |
1096 | static struct dentry *__lookup_hash(struct qstr *name, |
1097 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1098 | { |
057f6c01 | 1099 | struct dentry *dentry; |
1da177e4 LT |
1100 | struct inode *inode; |
1101 | int err; | |
1102 | ||
1103 | inode = base->d_inode; | |
1da177e4 LT |
1104 | |
1105 | /* | |
1106 | * See if the low-level filesystem might want | |
1107 | * to use its own hash.. | |
1108 | */ | |
1109 | if (base->d_op && base->d_op->d_hash) { | |
1110 | err = base->d_op->d_hash(base, name); | |
1111 | dentry = ERR_PTR(err); | |
1112 | if (err < 0) | |
1113 | goto out; | |
1114 | } | |
1115 | ||
6e6b1bd1 AV |
1116 | dentry = __d_lookup(base, name); |
1117 | ||
1118 | /* lockess __d_lookup may fail due to concurrent d_move() | |
1119 | * in some unrelated directory, so try with d_lookup | |
1120 | */ | |
1121 | if (!dentry) | |
1122 | dentry = d_lookup(base, name); | |
1123 | ||
1124 | if (dentry && dentry->d_op && dentry->d_op->d_revalidate) | |
1125 | dentry = do_revalidate(dentry, nd); | |
1126 | ||
1da177e4 | 1127 | if (!dentry) { |
d70b67c8 MS |
1128 | struct dentry *new; |
1129 | ||
1130 | /* Don't create child dentry for a dead directory. */ | |
1131 | dentry = ERR_PTR(-ENOENT); | |
1132 | if (IS_DEADDIR(inode)) | |
1133 | goto out; | |
1134 | ||
1135 | new = d_alloc(base, name); | |
1da177e4 LT |
1136 | dentry = ERR_PTR(-ENOMEM); |
1137 | if (!new) | |
1138 | goto out; | |
1139 | dentry = inode->i_op->lookup(inode, new, nd); | |
1140 | if (!dentry) | |
1141 | dentry = new; | |
1142 | else | |
1143 | dput(new); | |
1144 | } | |
1145 | out: | |
1146 | return dentry; | |
1147 | } | |
1148 | ||
057f6c01 JM |
1149 | /* |
1150 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1151 | * needs parent already locked. Doesn't follow mounts. | |
1152 | * SMP-safe. | |
1153 | */ | |
eead1911 | 1154 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1155 | { |
057f6c01 JM |
1156 | int err; |
1157 | ||
f419a2e3 | 1158 | err = inode_permission(nd->path.dentry->d_inode, MAY_EXEC); |
057f6c01 | 1159 | if (err) |
eead1911 | 1160 | return ERR_PTR(err); |
4ac91378 | 1161 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1162 | } |
1163 | ||
eead1911 CH |
1164 | static int __lookup_one_len(const char *name, struct qstr *this, |
1165 | struct dentry *base, int len) | |
1da177e4 LT |
1166 | { |
1167 | unsigned long hash; | |
1da177e4 LT |
1168 | unsigned int c; |
1169 | ||
057f6c01 JM |
1170 | this->name = name; |
1171 | this->len = len; | |
1da177e4 | 1172 | if (!len) |
057f6c01 | 1173 | return -EACCES; |
1da177e4 LT |
1174 | |
1175 | hash = init_name_hash(); | |
1176 | while (len--) { | |
1177 | c = *(const unsigned char *)name++; | |
1178 | if (c == '/' || c == '\0') | |
057f6c01 | 1179 | return -EACCES; |
1da177e4 LT |
1180 | hash = partial_name_hash(c, hash); |
1181 | } | |
057f6c01 JM |
1182 | this->hash = end_name_hash(hash); |
1183 | return 0; | |
1184 | } | |
1da177e4 | 1185 | |
eead1911 | 1186 | /** |
a6b91919 | 1187 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1188 | * @name: pathname component to lookup |
1189 | * @base: base directory to lookup from | |
1190 | * @len: maximum length @len should be interpreted to | |
1191 | * | |
a6b91919 RD |
1192 | * Note that this routine is purely a helper for filesystem usage and should |
1193 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1194 | * nameidata argument is passed to the filesystem methods and a filesystem |
1195 | * using this helper needs to be prepared for that. | |
1196 | */ | |
057f6c01 JM |
1197 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1198 | { | |
1199 | int err; | |
1200 | struct qstr this; | |
1201 | ||
2f9092e1 DW |
1202 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1203 | ||
057f6c01 | 1204 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1205 | if (err) |
1206 | return ERR_PTR(err); | |
1207 | ||
f419a2e3 | 1208 | err = inode_permission(base->d_inode, MAY_EXEC); |
057f6c01 JM |
1209 | if (err) |
1210 | return ERR_PTR(err); | |
49705b77 | 1211 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1212 | } |
1213 | ||
2d8f3038 AV |
1214 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1215 | struct path *path) | |
1da177e4 | 1216 | { |
2d8f3038 | 1217 | struct nameidata nd; |
1da177e4 LT |
1218 | char *tmp = getname(name); |
1219 | int err = PTR_ERR(tmp); | |
1da177e4 | 1220 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1221 | |
1222 | BUG_ON(flags & LOOKUP_PARENT); | |
1223 | ||
1224 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1225 | putname(tmp); |
2d8f3038 AV |
1226 | if (!err) |
1227 | *path = nd.path; | |
1da177e4 LT |
1228 | } |
1229 | return err; | |
1230 | } | |
1231 | ||
2ad94ae6 AV |
1232 | static int user_path_parent(int dfd, const char __user *path, |
1233 | struct nameidata *nd, char **name) | |
1234 | { | |
1235 | char *s = getname(path); | |
1236 | int error; | |
1237 | ||
1238 | if (IS_ERR(s)) | |
1239 | return PTR_ERR(s); | |
1240 | ||
1241 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1242 | if (error) | |
1243 | putname(s); | |
1244 | else | |
1245 | *name = s; | |
1246 | ||
1247 | return error; | |
1248 | } | |
1249 | ||
1da177e4 LT |
1250 | /* |
1251 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1252 | * minimal. | |
1253 | */ | |
1254 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1255 | { | |
da9592ed DH |
1256 | uid_t fsuid = current_fsuid(); |
1257 | ||
1da177e4 LT |
1258 | if (!(dir->i_mode & S_ISVTX)) |
1259 | return 0; | |
da9592ed | 1260 | if (inode->i_uid == fsuid) |
1da177e4 | 1261 | return 0; |
da9592ed | 1262 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1263 | return 0; |
1264 | return !capable(CAP_FOWNER); | |
1265 | } | |
1266 | ||
1267 | /* | |
1268 | * Check whether we can remove a link victim from directory dir, check | |
1269 | * whether the type of victim is right. | |
1270 | * 1. We can't do it if dir is read-only (done in permission()) | |
1271 | * 2. We should have write and exec permissions on dir | |
1272 | * 3. We can't remove anything from append-only dir | |
1273 | * 4. We can't do anything with immutable dir (done in permission()) | |
1274 | * 5. If the sticky bit on dir is set we should either | |
1275 | * a. be owner of dir, or | |
1276 | * b. be owner of victim, or | |
1277 | * c. have CAP_FOWNER capability | |
1278 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1279 | * links pointing to it. | |
1280 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1281 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1282 | * 9. We can't remove a root or mountpoint. | |
1283 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1284 | * nfs_async_unlink(). | |
1285 | */ | |
858119e1 | 1286 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1287 | { |
1288 | int error; | |
1289 | ||
1290 | if (!victim->d_inode) | |
1291 | return -ENOENT; | |
1292 | ||
1293 | BUG_ON(victim->d_parent->d_inode != dir); | |
5a190ae6 | 1294 | audit_inode_child(victim->d_name.name, victim, dir); |
1da177e4 | 1295 | |
f419a2e3 | 1296 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1297 | if (error) |
1298 | return error; | |
1299 | if (IS_APPEND(dir)) | |
1300 | return -EPERM; | |
1301 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1302 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1303 | return -EPERM; |
1304 | if (isdir) { | |
1305 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1306 | return -ENOTDIR; | |
1307 | if (IS_ROOT(victim)) | |
1308 | return -EBUSY; | |
1309 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1310 | return -EISDIR; | |
1311 | if (IS_DEADDIR(dir)) | |
1312 | return -ENOENT; | |
1313 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1314 | return -EBUSY; | |
1315 | return 0; | |
1316 | } | |
1317 | ||
1318 | /* Check whether we can create an object with dentry child in directory | |
1319 | * dir. | |
1320 | * 1. We can't do it if child already exists (open has special treatment for | |
1321 | * this case, but since we are inlined it's OK) | |
1322 | * 2. We can't do it if dir is read-only (done in permission()) | |
1323 | * 3. We should have write and exec permissions on dir | |
1324 | * 4. We can't do it if dir is immutable (done in permission()) | |
1325 | */ | |
a95164d9 | 1326 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1327 | { |
1328 | if (child->d_inode) | |
1329 | return -EEXIST; | |
1330 | if (IS_DEADDIR(dir)) | |
1331 | return -ENOENT; | |
f419a2e3 | 1332 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1333 | } |
1334 | ||
1335 | /* | |
1da177e4 LT |
1336 | * O_DIRECTORY translates into forcing a directory lookup. |
1337 | */ | |
1338 | static inline int lookup_flags(unsigned int f) | |
1339 | { | |
1340 | unsigned long retval = LOOKUP_FOLLOW; | |
1341 | ||
1342 | if (f & O_NOFOLLOW) | |
1343 | retval &= ~LOOKUP_FOLLOW; | |
1344 | ||
1da177e4 LT |
1345 | if (f & O_DIRECTORY) |
1346 | retval |= LOOKUP_DIRECTORY; | |
1347 | ||
1348 | return retval; | |
1349 | } | |
1350 | ||
1351 | /* | |
1352 | * p1 and p2 should be directories on the same fs. | |
1353 | */ | |
1354 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1355 | { | |
1356 | struct dentry *p; | |
1357 | ||
1358 | if (p1 == p2) { | |
f2eace23 | 1359 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1360 | return NULL; |
1361 | } | |
1362 | ||
a11f3a05 | 1363 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1364 | |
e2761a11 OH |
1365 | p = d_ancestor(p2, p1); |
1366 | if (p) { | |
1367 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1368 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1369 | return p; | |
1da177e4 LT |
1370 | } |
1371 | ||
e2761a11 OH |
1372 | p = d_ancestor(p1, p2); |
1373 | if (p) { | |
1374 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1375 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1376 | return p; | |
1da177e4 LT |
1377 | } |
1378 | ||
f2eace23 IM |
1379 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1380 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1381 | return NULL; |
1382 | } | |
1383 | ||
1384 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1385 | { | |
1b1dcc1b | 1386 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1387 | if (p1 != p2) { |
1b1dcc1b | 1388 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1389 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1390 | } |
1391 | } | |
1392 | ||
1393 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1394 | struct nameidata *nd) | |
1395 | { | |
a95164d9 | 1396 | int error = may_create(dir, dentry); |
1da177e4 LT |
1397 | |
1398 | if (error) | |
1399 | return error; | |
1400 | ||
acfa4380 | 1401 | if (!dir->i_op->create) |
1da177e4 LT |
1402 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1403 | mode &= S_IALLUGO; | |
1404 | mode |= S_IFREG; | |
1405 | error = security_inode_create(dir, dentry, mode); | |
1406 | if (error) | |
1407 | return error; | |
9e3509e2 | 1408 | vfs_dq_init(dir); |
1da177e4 | 1409 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1410 | if (!error) |
f38aa942 | 1411 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1412 | return error; |
1413 | } | |
1414 | ||
3fb64190 | 1415 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1416 | { |
3fb64190 | 1417 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1418 | struct inode *inode = dentry->d_inode; |
1419 | int error; | |
1420 | ||
1421 | if (!inode) | |
1422 | return -ENOENT; | |
1423 | ||
c8fe8f30 CH |
1424 | switch (inode->i_mode & S_IFMT) { |
1425 | case S_IFLNK: | |
1da177e4 | 1426 | return -ELOOP; |
c8fe8f30 CH |
1427 | case S_IFDIR: |
1428 | if (acc_mode & MAY_WRITE) | |
1429 | return -EISDIR; | |
1430 | break; | |
1431 | case S_IFBLK: | |
1432 | case S_IFCHR: | |
3fb64190 | 1433 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1434 | return -EACCES; |
c8fe8f30 CH |
1435 | /*FALLTHRU*/ |
1436 | case S_IFIFO: | |
1437 | case S_IFSOCK: | |
1da177e4 | 1438 | flag &= ~O_TRUNC; |
c8fe8f30 | 1439 | break; |
4a3fd211 | 1440 | } |
b41572e9 | 1441 | |
3fb64190 | 1442 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1443 | if (error) |
1444 | return error; | |
6146f0d5 | 1445 | |
6c1488fd MZ |
1446 | error = ima_path_check(path, acc_mode ? |
1447 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC) : | |
1448 | ACC_MODE(flag) & (MAY_READ | MAY_WRITE), | |
b9fc745d | 1449 | IMA_COUNT_UPDATE); |
6c1488fd | 1450 | |
6146f0d5 MZ |
1451 | if (error) |
1452 | return error; | |
1da177e4 LT |
1453 | /* |
1454 | * An append-only file must be opened in append mode for writing. | |
1455 | */ | |
1456 | if (IS_APPEND(inode)) { | |
acd0c935 | 1457 | error = -EPERM; |
1da177e4 | 1458 | if ((flag & FMODE_WRITE) && !(flag & O_APPEND)) |
acd0c935 | 1459 | goto err_out; |
1da177e4 | 1460 | if (flag & O_TRUNC) |
acd0c935 | 1461 | goto err_out; |
1da177e4 LT |
1462 | } |
1463 | ||
1464 | /* O_NOATIME can only be set by the owner or superuser */ | |
1465 | if (flag & O_NOATIME) | |
acd0c935 MZ |
1466 | if (!is_owner_or_cap(inode)) { |
1467 | error = -EPERM; | |
1468 | goto err_out; | |
1469 | } | |
1da177e4 LT |
1470 | |
1471 | /* | |
1472 | * Ensure there are no outstanding leases on the file. | |
1473 | */ | |
1474 | error = break_lease(inode, flag); | |
1475 | if (error) | |
acd0c935 | 1476 | goto err_out; |
1da177e4 LT |
1477 | |
1478 | if (flag & O_TRUNC) { | |
1479 | error = get_write_access(inode); | |
1480 | if (error) | |
acd0c935 | 1481 | goto err_out; |
1da177e4 LT |
1482 | |
1483 | /* | |
1484 | * Refuse to truncate files with mandatory locks held on them. | |
1485 | */ | |
1486 | error = locks_verify_locked(inode); | |
be6d3e56 | 1487 | if (!error) |
3fb64190 | 1488 | error = security_path_truncate(path, 0, |
be6d3e56 | 1489 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN); |
1da177e4 | 1490 | if (!error) { |
9e3509e2 | 1491 | vfs_dq_init(inode); |
d139d7ff MS |
1492 | |
1493 | error = do_truncate(dentry, 0, | |
1494 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1495 | NULL); | |
1da177e4 LT |
1496 | } |
1497 | put_write_access(inode); | |
1498 | if (error) | |
acd0c935 | 1499 | goto err_out; |
1da177e4 LT |
1500 | } else |
1501 | if (flag & FMODE_WRITE) | |
9e3509e2 | 1502 | vfs_dq_init(inode); |
1da177e4 LT |
1503 | |
1504 | return 0; | |
acd0c935 MZ |
1505 | err_out: |
1506 | ima_counts_put(path, acc_mode ? | |
1507 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC) : | |
1508 | ACC_MODE(flag) & (MAY_READ | MAY_WRITE)); | |
1509 | return error; | |
1da177e4 LT |
1510 | } |
1511 | ||
d57999e1 DH |
1512 | /* |
1513 | * Be careful about ever adding any more callers of this | |
1514 | * function. Its flags must be in the namei format, not | |
1515 | * what get passed to sys_open(). | |
1516 | */ | |
1517 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
aab520e2 DH |
1518 | int flag, int mode) |
1519 | { | |
1520 | int error; | |
4ac91378 | 1521 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1522 | |
1523 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1524 | mode &= ~current_umask(); |
be6d3e56 KT |
1525 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1526 | if (error) | |
1527 | goto out_unlock; | |
aab520e2 | 1528 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1529 | out_unlock: |
aab520e2 | 1530 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1531 | dput(nd->path.dentry); |
1532 | nd->path.dentry = path->dentry; | |
aab520e2 DH |
1533 | if (error) |
1534 | return error; | |
1535 | /* Don't check for write permission, don't truncate */ | |
3fb64190 | 1536 | return may_open(&nd->path, 0, flag & ~O_TRUNC); |
aab520e2 DH |
1537 | } |
1538 | ||
d57999e1 DH |
1539 | /* |
1540 | * Note that while the flag value (low two bits) for sys_open means: | |
1541 | * 00 - read-only | |
1542 | * 01 - write-only | |
1543 | * 10 - read-write | |
1544 | * 11 - special | |
1545 | * it is changed into | |
1546 | * 00 - no permissions needed | |
1547 | * 01 - read-permission | |
1548 | * 10 - write-permission | |
1549 | * 11 - read-write | |
1550 | * for the internal routines (ie open_namei()/follow_link() etc) | |
1551 | * This is more logical, and also allows the 00 "no perm needed" | |
1552 | * to be used for symlinks (where the permissions are checked | |
1553 | * later). | |
1554 | * | |
1555 | */ | |
1556 | static inline int open_to_namei_flags(int flag) | |
1557 | { | |
1558 | if ((flag+1) & O_ACCMODE) | |
1559 | flag++; | |
1560 | return flag; | |
1561 | } | |
1562 | ||
4a3fd211 DH |
1563 | static int open_will_write_to_fs(int flag, struct inode *inode) |
1564 | { | |
1565 | /* | |
1566 | * We'll never write to the fs underlying | |
1567 | * a device file. | |
1568 | */ | |
1569 | if (special_file(inode->i_mode)) | |
1570 | return 0; | |
1571 | return (flag & O_TRUNC); | |
1572 | } | |
1573 | ||
1da177e4 | 1574 | /* |
4a3fd211 DH |
1575 | * Note that the low bits of the passed in "open_flag" |
1576 | * are not the same as in the local variable "flag". See | |
1577 | * open_to_namei_flags() for more details. | |
1da177e4 | 1578 | */ |
a70e65df | 1579 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 1580 | int open_flag, int mode, int acc_mode) |
1da177e4 | 1581 | { |
4a3fd211 | 1582 | struct file *filp; |
a70e65df | 1583 | struct nameidata nd; |
6e8341a1 | 1584 | int error; |
6de88d72 | 1585 | struct path path, save; |
1da177e4 LT |
1586 | struct dentry *dir; |
1587 | int count = 0; | |
4a3fd211 | 1588 | int will_write; |
d57999e1 | 1589 | int flag = open_to_namei_flags(open_flag); |
1da177e4 | 1590 | |
6b2f3d1f CH |
1591 | /* |
1592 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1593 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1594 | * always set instead of having to deal with possibly weird behaviour | |
1595 | * for malicious applications setting only __O_SYNC. | |
1596 | */ | |
1597 | if (open_flag & __O_SYNC) | |
1598 | open_flag |= O_DSYNC; | |
1599 | ||
6e8341a1 AV |
1600 | if (!acc_mode) |
1601 | acc_mode = MAY_OPEN | ACC_MODE(flag); | |
1da177e4 | 1602 | |
834f2a4a TM |
1603 | /* O_TRUNC implies we need access checks for write permissions */ |
1604 | if (flag & O_TRUNC) | |
1605 | acc_mode |= MAY_WRITE; | |
1606 | ||
1da177e4 LT |
1607 | /* Allow the LSM permission hook to distinguish append |
1608 | access from general write access. */ | |
1609 | if (flag & O_APPEND) | |
1610 | acc_mode |= MAY_APPEND; | |
1611 | ||
1da177e4 LT |
1612 | /* |
1613 | * The simplest case - just a plain lookup. | |
1614 | */ | |
1615 | if (!(flag & O_CREAT)) { | |
2dd6d1f4 AV |
1616 | filp = get_empty_filp(); |
1617 | ||
1618 | if (filp == NULL) | |
1619 | return ERR_PTR(-ENFILE); | |
1620 | nd.intent.open.file = filp; | |
1621 | nd.intent.open.flags = flag; | |
1622 | nd.intent.open.create_mode = 0; | |
1623 | error = do_path_lookup(dfd, pathname, | |
1624 | lookup_flags(flag)|LOOKUP_OPEN, &nd); | |
1625 | if (IS_ERR(nd.intent.open.file)) { | |
1626 | if (error == 0) { | |
1627 | error = PTR_ERR(nd.intent.open.file); | |
1628 | path_put(&nd.path); | |
1629 | } | |
1630 | } else if (error) | |
1631 | release_open_intent(&nd); | |
1da177e4 | 1632 | if (error) |
a70e65df | 1633 | return ERR_PTR(error); |
1da177e4 LT |
1634 | goto ok; |
1635 | } | |
1636 | ||
1637 | /* | |
1638 | * Create - we need to know the parent. | |
1639 | */ | |
9b4a9b14 | 1640 | error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); |
1da177e4 | 1641 | if (error) |
a70e65df | 1642 | return ERR_PTR(error); |
9b4a9b14 | 1643 | error = path_walk(pathname, &nd); |
654f562c O |
1644 | if (error) { |
1645 | if (nd.root.mnt) | |
1646 | path_put(&nd.root); | |
9b4a9b14 | 1647 | return ERR_PTR(error); |
654f562c | 1648 | } |
9b4a9b14 AV |
1649 | if (unlikely(!audit_dummy_context())) |
1650 | audit_inode(pathname, nd.path.dentry); | |
1da177e4 LT |
1651 | |
1652 | /* | |
1653 | * We have the parent and last component. First of all, check | |
1654 | * that we are not asked to creat(2) an obvious directory - that | |
1655 | * will not do. | |
1656 | */ | |
1657 | error = -EISDIR; | |
a70e65df | 1658 | if (nd.last_type != LAST_NORM || nd.last.name[nd.last.len]) |
8737f3a1 | 1659 | goto exit_parent; |
1da177e4 | 1660 | |
8737f3a1 AV |
1661 | error = -ENFILE; |
1662 | filp = get_empty_filp(); | |
1663 | if (filp == NULL) | |
1664 | goto exit_parent; | |
1665 | nd.intent.open.file = filp; | |
1666 | nd.intent.open.flags = flag; | |
1667 | nd.intent.open.create_mode = mode; | |
a70e65df CH |
1668 | dir = nd.path.dentry; |
1669 | nd.flags &= ~LOOKUP_PARENT; | |
8737f3a1 | 1670 | nd.flags |= LOOKUP_CREATE | LOOKUP_OPEN; |
3516586a AV |
1671 | if (flag & O_EXCL) |
1672 | nd.flags |= LOOKUP_EXCL; | |
1b1dcc1b | 1673 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1674 | path.dentry = lookup_hash(&nd); |
1675 | path.mnt = nd.path.mnt; | |
1da177e4 LT |
1676 | |
1677 | do_last: | |
4e7506e4 AV |
1678 | error = PTR_ERR(path.dentry); |
1679 | if (IS_ERR(path.dentry)) { | |
1b1dcc1b | 1680 | mutex_unlock(&dir->d_inode->i_mutex); |
1da177e4 LT |
1681 | goto exit; |
1682 | } | |
1683 | ||
a70e65df | 1684 | if (IS_ERR(nd.intent.open.file)) { |
a70e65df | 1685 | error = PTR_ERR(nd.intent.open.file); |
4a3fd211 | 1686 | goto exit_mutex_unlock; |
4af4c52f OD |
1687 | } |
1688 | ||
1da177e4 | 1689 | /* Negative dentry, just create the file */ |
4e7506e4 | 1690 | if (!path.dentry->d_inode) { |
4a3fd211 DH |
1691 | /* |
1692 | * This write is needed to ensure that a | |
1693 | * ro->rw transition does not occur between | |
1694 | * the time when the file is created and when | |
1695 | * a permanent write count is taken through | |
1696 | * the 'struct file' in nameidata_to_filp(). | |
1697 | */ | |
1698 | error = mnt_want_write(nd.path.mnt); | |
1da177e4 | 1699 | if (error) |
4a3fd211 DH |
1700 | goto exit_mutex_unlock; |
1701 | error = __open_namei_create(&nd, &path, flag, mode); | |
1702 | if (error) { | |
1703 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1704 | goto exit; |
4a3fd211 DH |
1705 | } |
1706 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1707 | if (IS_ERR(filp)) |
1708 | ima_counts_put(&nd.path, | |
1709 | acc_mode & (MAY_READ | MAY_WRITE | | |
1710 | MAY_EXEC)); | |
4a3fd211 | 1711 | mnt_drop_write(nd.path.mnt); |
654f562c O |
1712 | if (nd.root.mnt) |
1713 | path_put(&nd.root); | |
4a3fd211 | 1714 | return filp; |
1da177e4 LT |
1715 | } |
1716 | ||
1717 | /* | |
1718 | * It already exists. | |
1719 | */ | |
1b1dcc1b | 1720 | mutex_unlock(&dir->d_inode->i_mutex); |
5a190ae6 | 1721 | audit_inode(pathname, path.dentry); |
1da177e4 LT |
1722 | |
1723 | error = -EEXIST; | |
1724 | if (flag & O_EXCL) | |
1725 | goto exit_dput; | |
1726 | ||
e13b210f | 1727 | if (__follow_mount(&path)) { |
1da177e4 | 1728 | error = -ELOOP; |
ba7a4c1a AV |
1729 | if (flag & O_NOFOLLOW) |
1730 | goto exit_dput; | |
1da177e4 | 1731 | } |
3e2efce0 | 1732 | |
1da177e4 | 1733 | error = -ENOENT; |
4e7506e4 | 1734 | if (!path.dentry->d_inode) |
1da177e4 | 1735 | goto exit_dput; |
acfa4380 | 1736 | if (path.dentry->d_inode->i_op->follow_link) |
1da177e4 LT |
1737 | goto do_link; |
1738 | ||
a70e65df | 1739 | path_to_nameidata(&path, &nd); |
1da177e4 | 1740 | error = -EISDIR; |
4e7506e4 | 1741 | if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode)) |
1da177e4 LT |
1742 | goto exit; |
1743 | ok: | |
4a3fd211 DH |
1744 | /* |
1745 | * Consider: | |
1746 | * 1. may_open() truncates a file | |
1747 | * 2. a rw->ro mount transition occurs | |
1748 | * 3. nameidata_to_filp() fails due to | |
1749 | * the ro mount. | |
1750 | * That would be inconsistent, and should | |
1751 | * be avoided. Taking this mnt write here | |
1752 | * ensures that (2) can not occur. | |
1753 | */ | |
1754 | will_write = open_will_write_to_fs(flag, nd.path.dentry->d_inode); | |
1755 | if (will_write) { | |
1756 | error = mnt_want_write(nd.path.mnt); | |
1757 | if (error) | |
1758 | goto exit; | |
1759 | } | |
3fb64190 | 1760 | error = may_open(&nd.path, acc_mode, flag); |
4a3fd211 DH |
1761 | if (error) { |
1762 | if (will_write) | |
1763 | mnt_drop_write(nd.path.mnt); | |
1da177e4 | 1764 | goto exit; |
4a3fd211 DH |
1765 | } |
1766 | filp = nameidata_to_filp(&nd, open_flag); | |
94e5d714 MZ |
1767 | if (IS_ERR(filp)) |
1768 | ima_counts_put(&nd.path, | |
1769 | acc_mode & (MAY_READ | MAY_WRITE | MAY_EXEC)); | |
4a3fd211 DH |
1770 | /* |
1771 | * It is now safe to drop the mnt write | |
1772 | * because the filp has had a write taken | |
1773 | * on its behalf. | |
1774 | */ | |
1775 | if (will_write) | |
1776 | mnt_drop_write(nd.path.mnt); | |
654f562c O |
1777 | if (nd.root.mnt) |
1778 | path_put(&nd.root); | |
4a3fd211 | 1779 | return filp; |
1da177e4 | 1780 | |
4a3fd211 DH |
1781 | exit_mutex_unlock: |
1782 | mutex_unlock(&dir->d_inode->i_mutex); | |
1da177e4 | 1783 | exit_dput: |
a70e65df | 1784 | path_put_conditional(&path, &nd); |
1da177e4 | 1785 | exit: |
a70e65df CH |
1786 | if (!IS_ERR(nd.intent.open.file)) |
1787 | release_open_intent(&nd); | |
8737f3a1 | 1788 | exit_parent: |
2a737871 AV |
1789 | if (nd.root.mnt) |
1790 | path_put(&nd.root); | |
a70e65df CH |
1791 | path_put(&nd.path); |
1792 | return ERR_PTR(error); | |
1da177e4 LT |
1793 | |
1794 | do_link: | |
1795 | error = -ELOOP; | |
1796 | if (flag & O_NOFOLLOW) | |
1797 | goto exit_dput; | |
1798 | /* | |
1799 | * This is subtle. Instead of calling do_follow_link() we do the | |
1800 | * thing by hands. The reason is that this way we have zero link_count | |
1801 | * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT. | |
1802 | * After that we have the parent and last component, i.e. | |
1803 | * we are in the same situation as after the first path_walk(). | |
1804 | * Well, almost - if the last component is normal we get its copy | |
1805 | * stored in nd->last.name and we will have to putname() it when we | |
1806 | * are done. Procfs-like symlinks just set LAST_BIND. | |
1807 | */ | |
a70e65df CH |
1808 | nd.flags |= LOOKUP_PARENT; |
1809 | error = security_inode_follow_link(path.dentry, &nd); | |
1da177e4 LT |
1810 | if (error) |
1811 | goto exit_dput; | |
6de88d72 AV |
1812 | save = nd.path; |
1813 | path_get(&save); | |
a70e65df | 1814 | error = __do_follow_link(&path, &nd); |
6de88d72 AV |
1815 | if (error == -ESTALE) { |
1816 | /* nd.path had been dropped */ | |
1817 | nd.path = save; | |
1818 | path_get(&nd.path); | |
1819 | nd.flags |= LOOKUP_REVAL; | |
1820 | error = __do_follow_link(&path, &nd); | |
1821 | } | |
1822 | path_put(&save); | |
258fa999 | 1823 | path_put(&path); |
de459215 KK |
1824 | if (error) { |
1825 | /* Does someone understand code flow here? Or it is only | |
1826 | * me so stupid? Anathema to whoever designed this non-sense | |
1827 | * with "intent.open". | |
1828 | */ | |
a70e65df | 1829 | release_open_intent(&nd); |
654f562c O |
1830 | if (nd.root.mnt) |
1831 | path_put(&nd.root); | |
a70e65df | 1832 | return ERR_PTR(error); |
de459215 | 1833 | } |
a70e65df CH |
1834 | nd.flags &= ~LOOKUP_PARENT; |
1835 | if (nd.last_type == LAST_BIND) | |
1da177e4 | 1836 | goto ok; |
1da177e4 | 1837 | error = -EISDIR; |
a70e65df | 1838 | if (nd.last_type != LAST_NORM) |
1da177e4 | 1839 | goto exit; |
a70e65df CH |
1840 | if (nd.last.name[nd.last.len]) { |
1841 | __putname(nd.last.name); | |
1da177e4 LT |
1842 | goto exit; |
1843 | } | |
1844 | error = -ELOOP; | |
1845 | if (count++==32) { | |
a70e65df | 1846 | __putname(nd.last.name); |
1da177e4 LT |
1847 | goto exit; |
1848 | } | |
a70e65df | 1849 | dir = nd.path.dentry; |
1b1dcc1b | 1850 | mutex_lock(&dir->d_inode->i_mutex); |
a70e65df CH |
1851 | path.dentry = lookup_hash(&nd); |
1852 | path.mnt = nd.path.mnt; | |
1853 | __putname(nd.last.name); | |
1da177e4 LT |
1854 | goto do_last; |
1855 | } | |
1856 | ||
a70e65df CH |
1857 | /** |
1858 | * filp_open - open file and return file pointer | |
1859 | * | |
1860 | * @filename: path to open | |
1861 | * @flags: open flags as per the open(2) second argument | |
1862 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1863 | * | |
1864 | * This is the helper to open a file from kernelspace if you really | |
1865 | * have to. But in generally you should not do this, so please move | |
1866 | * along, nothing to see here.. | |
1867 | */ | |
1868 | struct file *filp_open(const char *filename, int flags, int mode) | |
1869 | { | |
6e8341a1 | 1870 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
1871 | } |
1872 | EXPORT_SYMBOL(filp_open); | |
1873 | ||
1da177e4 LT |
1874 | /** |
1875 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
1876 | * @nd: nameidata info | |
1877 | * @is_dir: directory flag | |
1878 | * | |
1879 | * Simple function to lookup and return a dentry and create it | |
1880 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 1881 | * |
4ac91378 | 1882 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
1883 | */ |
1884 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
1885 | { | |
c663e5d8 | 1886 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 1887 | |
4ac91378 | 1888 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
1889 | /* |
1890 | * Yucky last component or no last component at all? | |
1891 | * (foo/., foo/.., /////) | |
1892 | */ | |
1da177e4 LT |
1893 | if (nd->last_type != LAST_NORM) |
1894 | goto fail; | |
1895 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 1896 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 1897 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
1898 | |
1899 | /* | |
1900 | * Do the final lookup. | |
1901 | */ | |
49705b77 | 1902 | dentry = lookup_hash(nd); |
1da177e4 LT |
1903 | if (IS_ERR(dentry)) |
1904 | goto fail; | |
c663e5d8 | 1905 | |
e9baf6e5 AV |
1906 | if (dentry->d_inode) |
1907 | goto eexist; | |
c663e5d8 CH |
1908 | /* |
1909 | * Special case - lookup gave negative, but... we had foo/bar/ | |
1910 | * From the vfs_mknod() POV we just have a negative dentry - | |
1911 | * all is fine. Let's be bastards - you had / on the end, you've | |
1912 | * been asking for (non-existent) directory. -ENOENT for you. | |
1913 | */ | |
e9baf6e5 AV |
1914 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
1915 | dput(dentry); | |
1916 | dentry = ERR_PTR(-ENOENT); | |
1917 | } | |
1da177e4 | 1918 | return dentry; |
e9baf6e5 | 1919 | eexist: |
1da177e4 | 1920 | dput(dentry); |
e9baf6e5 | 1921 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
1922 | fail: |
1923 | return dentry; | |
1924 | } | |
f81a0bff | 1925 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
1926 | |
1927 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
1928 | { | |
a95164d9 | 1929 | int error = may_create(dir, dentry); |
1da177e4 LT |
1930 | |
1931 | if (error) | |
1932 | return error; | |
1933 | ||
1934 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
1935 | return -EPERM; | |
1936 | ||
acfa4380 | 1937 | if (!dir->i_op->mknod) |
1da177e4 LT |
1938 | return -EPERM; |
1939 | ||
08ce5f16 SH |
1940 | error = devcgroup_inode_mknod(mode, dev); |
1941 | if (error) | |
1942 | return error; | |
1943 | ||
1da177e4 LT |
1944 | error = security_inode_mknod(dir, dentry, mode, dev); |
1945 | if (error) | |
1946 | return error; | |
1947 | ||
9e3509e2 | 1948 | vfs_dq_init(dir); |
1da177e4 | 1949 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 1950 | if (!error) |
f38aa942 | 1951 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1952 | return error; |
1953 | } | |
1954 | ||
463c3197 DH |
1955 | static int may_mknod(mode_t mode) |
1956 | { | |
1957 | switch (mode & S_IFMT) { | |
1958 | case S_IFREG: | |
1959 | case S_IFCHR: | |
1960 | case S_IFBLK: | |
1961 | case S_IFIFO: | |
1962 | case S_IFSOCK: | |
1963 | case 0: /* zero mode translates to S_IFREG */ | |
1964 | return 0; | |
1965 | case S_IFDIR: | |
1966 | return -EPERM; | |
1967 | default: | |
1968 | return -EINVAL; | |
1969 | } | |
1970 | } | |
1971 | ||
2e4d0924 HC |
1972 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
1973 | unsigned, dev) | |
1da177e4 | 1974 | { |
2ad94ae6 AV |
1975 | int error; |
1976 | char *tmp; | |
1977 | struct dentry *dentry; | |
1da177e4 LT |
1978 | struct nameidata nd; |
1979 | ||
1980 | if (S_ISDIR(mode)) | |
1981 | return -EPERM; | |
1da177e4 | 1982 | |
2ad94ae6 | 1983 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 1984 | if (error) |
2ad94ae6 AV |
1985 | return error; |
1986 | ||
1da177e4 | 1987 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
1988 | if (IS_ERR(dentry)) { |
1989 | error = PTR_ERR(dentry); | |
1990 | goto out_unlock; | |
1991 | } | |
4ac91378 | 1992 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 1993 | mode &= ~current_umask(); |
463c3197 DH |
1994 | error = may_mknod(mode); |
1995 | if (error) | |
1996 | goto out_dput; | |
1997 | error = mnt_want_write(nd.path.mnt); | |
1998 | if (error) | |
1999 | goto out_dput; | |
be6d3e56 KT |
2000 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2001 | if (error) | |
2002 | goto out_drop_write; | |
463c3197 | 2003 | switch (mode & S_IFMT) { |
1da177e4 | 2004 | case 0: case S_IFREG: |
4ac91378 | 2005 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2006 | break; |
2007 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2008 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2009 | new_decode_dev(dev)); |
2010 | break; | |
2011 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2012 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2013 | break; |
1da177e4 | 2014 | } |
be6d3e56 | 2015 | out_drop_write: |
463c3197 DH |
2016 | mnt_drop_write(nd.path.mnt); |
2017 | out_dput: | |
2018 | dput(dentry); | |
2019 | out_unlock: | |
4ac91378 | 2020 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2021 | path_put(&nd.path); |
1da177e4 LT |
2022 | putname(tmp); |
2023 | ||
2024 | return error; | |
2025 | } | |
2026 | ||
3480b257 | 2027 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2028 | { |
2029 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2030 | } | |
2031 | ||
1da177e4 LT |
2032 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2033 | { | |
a95164d9 | 2034 | int error = may_create(dir, dentry); |
1da177e4 LT |
2035 | |
2036 | if (error) | |
2037 | return error; | |
2038 | ||
acfa4380 | 2039 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2040 | return -EPERM; |
2041 | ||
2042 | mode &= (S_IRWXUGO|S_ISVTX); | |
2043 | error = security_inode_mkdir(dir, dentry, mode); | |
2044 | if (error) | |
2045 | return error; | |
2046 | ||
9e3509e2 | 2047 | vfs_dq_init(dir); |
1da177e4 | 2048 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2049 | if (!error) |
f38aa942 | 2050 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2051 | return error; |
2052 | } | |
2053 | ||
2e4d0924 | 2054 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2055 | { |
2056 | int error = 0; | |
2057 | char * tmp; | |
6902d925 DH |
2058 | struct dentry *dentry; |
2059 | struct nameidata nd; | |
1da177e4 | 2060 | |
2ad94ae6 AV |
2061 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2062 | if (error) | |
6902d925 | 2063 | goto out_err; |
1da177e4 | 2064 | |
6902d925 DH |
2065 | dentry = lookup_create(&nd, 1); |
2066 | error = PTR_ERR(dentry); | |
2067 | if (IS_ERR(dentry)) | |
2068 | goto out_unlock; | |
1da177e4 | 2069 | |
4ac91378 | 2070 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2071 | mode &= ~current_umask(); |
463c3197 DH |
2072 | error = mnt_want_write(nd.path.mnt); |
2073 | if (error) | |
2074 | goto out_dput; | |
be6d3e56 KT |
2075 | error = security_path_mkdir(&nd.path, dentry, mode); |
2076 | if (error) | |
2077 | goto out_drop_write; | |
4ac91378 | 2078 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2079 | out_drop_write: |
463c3197 DH |
2080 | mnt_drop_write(nd.path.mnt); |
2081 | out_dput: | |
6902d925 DH |
2082 | dput(dentry); |
2083 | out_unlock: | |
4ac91378 | 2084 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2085 | path_put(&nd.path); |
6902d925 DH |
2086 | putname(tmp); |
2087 | out_err: | |
1da177e4 LT |
2088 | return error; |
2089 | } | |
2090 | ||
3cdad428 | 2091 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2092 | { |
2093 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2094 | } | |
2095 | ||
1da177e4 LT |
2096 | /* |
2097 | * We try to drop the dentry early: we should have | |
2098 | * a usage count of 2 if we're the only user of this | |
2099 | * dentry, and if that is true (possibly after pruning | |
2100 | * the dcache), then we drop the dentry now. | |
2101 | * | |
2102 | * A low-level filesystem can, if it choses, legally | |
2103 | * do a | |
2104 | * | |
2105 | * if (!d_unhashed(dentry)) | |
2106 | * return -EBUSY; | |
2107 | * | |
2108 | * if it cannot handle the case of removing a directory | |
2109 | * that is still in use by something else.. | |
2110 | */ | |
2111 | void dentry_unhash(struct dentry *dentry) | |
2112 | { | |
2113 | dget(dentry); | |
dc168427 | 2114 | shrink_dcache_parent(dentry); |
1da177e4 LT |
2115 | spin_lock(&dcache_lock); |
2116 | spin_lock(&dentry->d_lock); | |
2117 | if (atomic_read(&dentry->d_count) == 2) | |
2118 | __d_drop(dentry); | |
2119 | spin_unlock(&dentry->d_lock); | |
2120 | spin_unlock(&dcache_lock); | |
2121 | } | |
2122 | ||
2123 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2124 | { | |
2125 | int error = may_delete(dir, dentry, 1); | |
2126 | ||
2127 | if (error) | |
2128 | return error; | |
2129 | ||
acfa4380 | 2130 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2131 | return -EPERM; |
2132 | ||
9e3509e2 | 2133 | vfs_dq_init(dir); |
1da177e4 | 2134 | |
1b1dcc1b | 2135 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2136 | dentry_unhash(dentry); |
2137 | if (d_mountpoint(dentry)) | |
2138 | error = -EBUSY; | |
2139 | else { | |
2140 | error = security_inode_rmdir(dir, dentry); | |
2141 | if (!error) { | |
2142 | error = dir->i_op->rmdir(dir, dentry); | |
2143 | if (!error) | |
2144 | dentry->d_inode->i_flags |= S_DEAD; | |
2145 | } | |
2146 | } | |
1b1dcc1b | 2147 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2148 | if (!error) { |
1da177e4 LT |
2149 | d_delete(dentry); |
2150 | } | |
2151 | dput(dentry); | |
2152 | ||
2153 | return error; | |
2154 | } | |
2155 | ||
5590ff0d | 2156 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2157 | { |
2158 | int error = 0; | |
2159 | char * name; | |
2160 | struct dentry *dentry; | |
2161 | struct nameidata nd; | |
2162 | ||
2ad94ae6 | 2163 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2164 | if (error) |
2ad94ae6 | 2165 | return error; |
1da177e4 LT |
2166 | |
2167 | switch(nd.last_type) { | |
0612d9fb OH |
2168 | case LAST_DOTDOT: |
2169 | error = -ENOTEMPTY; | |
2170 | goto exit1; | |
2171 | case LAST_DOT: | |
2172 | error = -EINVAL; | |
2173 | goto exit1; | |
2174 | case LAST_ROOT: | |
2175 | error = -EBUSY; | |
2176 | goto exit1; | |
1da177e4 | 2177 | } |
0612d9fb OH |
2178 | |
2179 | nd.flags &= ~LOOKUP_PARENT; | |
2180 | ||
4ac91378 | 2181 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2182 | dentry = lookup_hash(&nd); |
1da177e4 | 2183 | error = PTR_ERR(dentry); |
6902d925 DH |
2184 | if (IS_ERR(dentry)) |
2185 | goto exit2; | |
0622753b DH |
2186 | error = mnt_want_write(nd.path.mnt); |
2187 | if (error) | |
2188 | goto exit3; | |
be6d3e56 KT |
2189 | error = security_path_rmdir(&nd.path, dentry); |
2190 | if (error) | |
2191 | goto exit4; | |
4ac91378 | 2192 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2193 | exit4: |
0622753b DH |
2194 | mnt_drop_write(nd.path.mnt); |
2195 | exit3: | |
6902d925 DH |
2196 | dput(dentry); |
2197 | exit2: | |
4ac91378 | 2198 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2199 | exit1: |
1d957f9b | 2200 | path_put(&nd.path); |
1da177e4 LT |
2201 | putname(name); |
2202 | return error; | |
2203 | } | |
2204 | ||
3cdad428 | 2205 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2206 | { |
2207 | return do_rmdir(AT_FDCWD, pathname); | |
2208 | } | |
2209 | ||
1da177e4 LT |
2210 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2211 | { | |
2212 | int error = may_delete(dir, dentry, 0); | |
2213 | ||
2214 | if (error) | |
2215 | return error; | |
2216 | ||
acfa4380 | 2217 | if (!dir->i_op->unlink) |
1da177e4 LT |
2218 | return -EPERM; |
2219 | ||
9e3509e2 | 2220 | vfs_dq_init(dir); |
1da177e4 | 2221 | |
1b1dcc1b | 2222 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2223 | if (d_mountpoint(dentry)) |
2224 | error = -EBUSY; | |
2225 | else { | |
2226 | error = security_inode_unlink(dir, dentry); | |
2227 | if (!error) | |
2228 | error = dir->i_op->unlink(dir, dentry); | |
2229 | } | |
1b1dcc1b | 2230 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2231 | |
2232 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2233 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2234 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2235 | d_delete(dentry); |
1da177e4 | 2236 | } |
0eeca283 | 2237 | |
1da177e4 LT |
2238 | return error; |
2239 | } | |
2240 | ||
2241 | /* | |
2242 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2243 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2244 | * writeout happening, and we don't want to prevent access to the directory |
2245 | * while waiting on the I/O. | |
2246 | */ | |
5590ff0d | 2247 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2248 | { |
2ad94ae6 AV |
2249 | int error; |
2250 | char *name; | |
1da177e4 LT |
2251 | struct dentry *dentry; |
2252 | struct nameidata nd; | |
2253 | struct inode *inode = NULL; | |
2254 | ||
2ad94ae6 | 2255 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2256 | if (error) |
2ad94ae6 AV |
2257 | return error; |
2258 | ||
1da177e4 LT |
2259 | error = -EISDIR; |
2260 | if (nd.last_type != LAST_NORM) | |
2261 | goto exit1; | |
0612d9fb OH |
2262 | |
2263 | nd.flags &= ~LOOKUP_PARENT; | |
2264 | ||
4ac91378 | 2265 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2266 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2267 | error = PTR_ERR(dentry); |
2268 | if (!IS_ERR(dentry)) { | |
2269 | /* Why not before? Because we want correct error value */ | |
2270 | if (nd.last.name[nd.last.len]) | |
2271 | goto slashes; | |
2272 | inode = dentry->d_inode; | |
2273 | if (inode) | |
2274 | atomic_inc(&inode->i_count); | |
0622753b DH |
2275 | error = mnt_want_write(nd.path.mnt); |
2276 | if (error) | |
2277 | goto exit2; | |
be6d3e56 KT |
2278 | error = security_path_unlink(&nd.path, dentry); |
2279 | if (error) | |
2280 | goto exit3; | |
4ac91378 | 2281 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2282 | exit3: |
0622753b | 2283 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2284 | exit2: |
2285 | dput(dentry); | |
2286 | } | |
4ac91378 | 2287 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2288 | if (inode) |
2289 | iput(inode); /* truncate the inode here */ | |
2290 | exit1: | |
1d957f9b | 2291 | path_put(&nd.path); |
1da177e4 LT |
2292 | putname(name); |
2293 | return error; | |
2294 | ||
2295 | slashes: | |
2296 | error = !dentry->d_inode ? -ENOENT : | |
2297 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2298 | goto exit2; | |
2299 | } | |
2300 | ||
2e4d0924 | 2301 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2302 | { |
2303 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2304 | return -EINVAL; | |
2305 | ||
2306 | if (flag & AT_REMOVEDIR) | |
2307 | return do_rmdir(dfd, pathname); | |
2308 | ||
2309 | return do_unlinkat(dfd, pathname); | |
2310 | } | |
2311 | ||
3480b257 | 2312 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2313 | { |
2314 | return do_unlinkat(AT_FDCWD, pathname); | |
2315 | } | |
2316 | ||
db2e747b | 2317 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2318 | { |
a95164d9 | 2319 | int error = may_create(dir, dentry); |
1da177e4 LT |
2320 | |
2321 | if (error) | |
2322 | return error; | |
2323 | ||
acfa4380 | 2324 | if (!dir->i_op->symlink) |
1da177e4 LT |
2325 | return -EPERM; |
2326 | ||
2327 | error = security_inode_symlink(dir, dentry, oldname); | |
2328 | if (error) | |
2329 | return error; | |
2330 | ||
9e3509e2 | 2331 | vfs_dq_init(dir); |
1da177e4 | 2332 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2333 | if (!error) |
f38aa942 | 2334 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2335 | return error; |
2336 | } | |
2337 | ||
2e4d0924 HC |
2338 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2339 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2340 | { |
2ad94ae6 AV |
2341 | int error; |
2342 | char *from; | |
2343 | char *to; | |
6902d925 DH |
2344 | struct dentry *dentry; |
2345 | struct nameidata nd; | |
1da177e4 LT |
2346 | |
2347 | from = getname(oldname); | |
2ad94ae6 | 2348 | if (IS_ERR(from)) |
1da177e4 | 2349 | return PTR_ERR(from); |
1da177e4 | 2350 | |
2ad94ae6 | 2351 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2352 | if (error) |
2ad94ae6 AV |
2353 | goto out_putname; |
2354 | ||
6902d925 DH |
2355 | dentry = lookup_create(&nd, 0); |
2356 | error = PTR_ERR(dentry); | |
2357 | if (IS_ERR(dentry)) | |
2358 | goto out_unlock; | |
2359 | ||
75c3f29d DH |
2360 | error = mnt_want_write(nd.path.mnt); |
2361 | if (error) | |
2362 | goto out_dput; | |
be6d3e56 KT |
2363 | error = security_path_symlink(&nd.path, dentry, from); |
2364 | if (error) | |
2365 | goto out_drop_write; | |
db2e747b | 2366 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2367 | out_drop_write: |
75c3f29d DH |
2368 | mnt_drop_write(nd.path.mnt); |
2369 | out_dput: | |
6902d925 DH |
2370 | dput(dentry); |
2371 | out_unlock: | |
4ac91378 | 2372 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2373 | path_put(&nd.path); |
6902d925 DH |
2374 | putname(to); |
2375 | out_putname: | |
1da177e4 LT |
2376 | putname(from); |
2377 | return error; | |
2378 | } | |
2379 | ||
3480b257 | 2380 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2381 | { |
2382 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2383 | } | |
2384 | ||
1da177e4 LT |
2385 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2386 | { | |
2387 | struct inode *inode = old_dentry->d_inode; | |
2388 | int error; | |
2389 | ||
2390 | if (!inode) | |
2391 | return -ENOENT; | |
2392 | ||
a95164d9 | 2393 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2394 | if (error) |
2395 | return error; | |
2396 | ||
2397 | if (dir->i_sb != inode->i_sb) | |
2398 | return -EXDEV; | |
2399 | ||
2400 | /* | |
2401 | * A link to an append-only or immutable file cannot be created. | |
2402 | */ | |
2403 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2404 | return -EPERM; | |
acfa4380 | 2405 | if (!dir->i_op->link) |
1da177e4 | 2406 | return -EPERM; |
7e79eedb | 2407 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2408 | return -EPERM; |
2409 | ||
2410 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2411 | if (error) | |
2412 | return error; | |
2413 | ||
7e79eedb | 2414 | mutex_lock(&inode->i_mutex); |
9e3509e2 | 2415 | vfs_dq_init(dir); |
1da177e4 | 2416 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2417 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2418 | if (!error) |
7e79eedb | 2419 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2420 | return error; |
2421 | } | |
2422 | ||
2423 | /* | |
2424 | * Hardlinks are often used in delicate situations. We avoid | |
2425 | * security-related surprises by not following symlinks on the | |
2426 | * newname. --KAB | |
2427 | * | |
2428 | * We don't follow them on the oldname either to be compatible | |
2429 | * with linux 2.0, and to avoid hard-linking to directories | |
2430 | * and other special files. --ADM | |
2431 | */ | |
2e4d0924 HC |
2432 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2433 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2434 | { |
2435 | struct dentry *new_dentry; | |
2d8f3038 AV |
2436 | struct nameidata nd; |
2437 | struct path old_path; | |
1da177e4 | 2438 | int error; |
2ad94ae6 | 2439 | char *to; |
1da177e4 | 2440 | |
45c9b11a | 2441 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2442 | return -EINVAL; |
2443 | ||
2d8f3038 AV |
2444 | error = user_path_at(olddfd, oldname, |
2445 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2446 | &old_path); | |
1da177e4 | 2447 | if (error) |
2ad94ae6 AV |
2448 | return error; |
2449 | ||
2450 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2451 | if (error) |
2452 | goto out; | |
2453 | error = -EXDEV; | |
2d8f3038 | 2454 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2455 | goto out_release; |
2456 | new_dentry = lookup_create(&nd, 0); | |
2457 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2458 | if (IS_ERR(new_dentry)) |
2459 | goto out_unlock; | |
75c3f29d DH |
2460 | error = mnt_want_write(nd.path.mnt); |
2461 | if (error) | |
2462 | goto out_dput; | |
be6d3e56 KT |
2463 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2464 | if (error) | |
2465 | goto out_drop_write; | |
2d8f3038 | 2466 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2467 | out_drop_write: |
75c3f29d DH |
2468 | mnt_drop_write(nd.path.mnt); |
2469 | out_dput: | |
6902d925 DH |
2470 | dput(new_dentry); |
2471 | out_unlock: | |
4ac91378 | 2472 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2473 | out_release: |
1d957f9b | 2474 | path_put(&nd.path); |
2ad94ae6 | 2475 | putname(to); |
1da177e4 | 2476 | out: |
2d8f3038 | 2477 | path_put(&old_path); |
1da177e4 LT |
2478 | |
2479 | return error; | |
2480 | } | |
2481 | ||
3480b257 | 2482 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2483 | { |
c04030e1 | 2484 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2485 | } |
2486 | ||
1da177e4 LT |
2487 | /* |
2488 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2489 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2490 | * Problems: | |
2491 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2492 | * b) race potential - two innocent renames can create a loop together. | |
2493 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2494 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2495 | * story. |
2496 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2497 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2498 | * whether the target exists). Solution: try to be smart with locking |
2499 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2500 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2501 | * move will be locked. Thus we can rank directories by the tree |
2502 | * (ancestors first) and rank all non-directories after them. | |
2503 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2504 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2505 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2506 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2507 | * we'd better make sure that there's no link(2) for them. | |
2508 | * d) some filesystems don't support opened-but-unlinked directories, | |
2509 | * either because of layout or because they are not ready to deal with | |
2510 | * all cases correctly. The latter will be fixed (taking this sort of | |
2511 | * stuff into VFS), but the former is not going away. Solution: the same | |
2512 | * trick as in rmdir(). | |
2513 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 2514 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2515 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
1b1dcc1b | 2516 | * ->i_mutex on parents, which works but leads to some truely excessive |
1da177e4 LT |
2517 | * locking]. |
2518 | */ | |
75c96f85 AB |
2519 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2520 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2521 | { |
2522 | int error = 0; | |
2523 | struct inode *target; | |
2524 | ||
2525 | /* | |
2526 | * If we are going to change the parent - check write permissions, | |
2527 | * we'll need to flip '..'. | |
2528 | */ | |
2529 | if (new_dir != old_dir) { | |
f419a2e3 | 2530 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2531 | if (error) |
2532 | return error; | |
2533 | } | |
2534 | ||
2535 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2536 | if (error) | |
2537 | return error; | |
2538 | ||
2539 | target = new_dentry->d_inode; | |
2540 | if (target) { | |
1b1dcc1b | 2541 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2542 | dentry_unhash(new_dentry); |
2543 | } | |
2544 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) | |
2545 | error = -EBUSY; | |
2546 | else | |
2547 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2548 | if (target) { | |
2549 | if (!error) | |
2550 | target->i_flags |= S_DEAD; | |
1b1dcc1b | 2551 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2552 | if (d_unhashed(new_dentry)) |
2553 | d_rehash(new_dentry); | |
2554 | dput(new_dentry); | |
2555 | } | |
e31e14ec | 2556 | if (!error) |
349457cc MF |
2557 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
2558 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
2559 | return error; |
2560 | } | |
2561 | ||
75c96f85 AB |
2562 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
2563 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2564 | { |
2565 | struct inode *target; | |
2566 | int error; | |
2567 | ||
2568 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2569 | if (error) | |
2570 | return error; | |
2571 | ||
2572 | dget(new_dentry); | |
2573 | target = new_dentry->d_inode; | |
2574 | if (target) | |
1b1dcc1b | 2575 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
2576 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
2577 | error = -EBUSY; | |
2578 | else | |
2579 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
2580 | if (!error) { | |
349457cc | 2581 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 2582 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
2583 | } |
2584 | if (target) | |
1b1dcc1b | 2585 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
2586 | dput(new_dentry); |
2587 | return error; | |
2588 | } | |
2589 | ||
2590 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
2591 | struct inode *new_dir, struct dentry *new_dentry) | |
2592 | { | |
2593 | int error; | |
2594 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
0eeca283 | 2595 | const char *old_name; |
1da177e4 LT |
2596 | |
2597 | if (old_dentry->d_inode == new_dentry->d_inode) | |
2598 | return 0; | |
2599 | ||
2600 | error = may_delete(old_dir, old_dentry, is_dir); | |
2601 | if (error) | |
2602 | return error; | |
2603 | ||
2604 | if (!new_dentry->d_inode) | |
a95164d9 | 2605 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
2606 | else |
2607 | error = may_delete(new_dir, new_dentry, is_dir); | |
2608 | if (error) | |
2609 | return error; | |
2610 | ||
acfa4380 | 2611 | if (!old_dir->i_op->rename) |
1da177e4 LT |
2612 | return -EPERM; |
2613 | ||
9e3509e2 JK |
2614 | vfs_dq_init(old_dir); |
2615 | vfs_dq_init(new_dir); | |
1da177e4 | 2616 | |
0eeca283 RL |
2617 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
2618 | ||
1da177e4 LT |
2619 | if (is_dir) |
2620 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
2621 | else | |
2622 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
2623 | if (!error) { | |
0eeca283 | 2624 | const char *new_name = old_dentry->d_name.name; |
89204c40 | 2625 | fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir, |
5a190ae6 | 2626 | new_dentry->d_inode, old_dentry); |
1da177e4 | 2627 | } |
0eeca283 RL |
2628 | fsnotify_oldname_free(old_name); |
2629 | ||
1da177e4 LT |
2630 | return error; |
2631 | } | |
2632 | ||
2e4d0924 HC |
2633 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
2634 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2635 | { |
2ad94ae6 AV |
2636 | struct dentry *old_dir, *new_dir; |
2637 | struct dentry *old_dentry, *new_dentry; | |
2638 | struct dentry *trap; | |
1da177e4 | 2639 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
2640 | char *from; |
2641 | char *to; | |
2642 | int error; | |
1da177e4 | 2643 | |
2ad94ae6 | 2644 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
2645 | if (error) |
2646 | goto exit; | |
2647 | ||
2ad94ae6 | 2648 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
2649 | if (error) |
2650 | goto exit1; | |
2651 | ||
2652 | error = -EXDEV; | |
4ac91378 | 2653 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
2654 | goto exit2; |
2655 | ||
4ac91378 | 2656 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
2657 | error = -EBUSY; |
2658 | if (oldnd.last_type != LAST_NORM) | |
2659 | goto exit2; | |
2660 | ||
4ac91378 | 2661 | new_dir = newnd.path.dentry; |
1da177e4 LT |
2662 | if (newnd.last_type != LAST_NORM) |
2663 | goto exit2; | |
2664 | ||
0612d9fb OH |
2665 | oldnd.flags &= ~LOOKUP_PARENT; |
2666 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 2667 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 2668 | |
1da177e4 LT |
2669 | trap = lock_rename(new_dir, old_dir); |
2670 | ||
49705b77 | 2671 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
2672 | error = PTR_ERR(old_dentry); |
2673 | if (IS_ERR(old_dentry)) | |
2674 | goto exit3; | |
2675 | /* source must exist */ | |
2676 | error = -ENOENT; | |
2677 | if (!old_dentry->d_inode) | |
2678 | goto exit4; | |
2679 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
2680 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
2681 | error = -ENOTDIR; | |
2682 | if (oldnd.last.name[oldnd.last.len]) | |
2683 | goto exit4; | |
2684 | if (newnd.last.name[newnd.last.len]) | |
2685 | goto exit4; | |
2686 | } | |
2687 | /* source should not be ancestor of target */ | |
2688 | error = -EINVAL; | |
2689 | if (old_dentry == trap) | |
2690 | goto exit4; | |
49705b77 | 2691 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
2692 | error = PTR_ERR(new_dentry); |
2693 | if (IS_ERR(new_dentry)) | |
2694 | goto exit4; | |
2695 | /* target should not be an ancestor of source */ | |
2696 | error = -ENOTEMPTY; | |
2697 | if (new_dentry == trap) | |
2698 | goto exit5; | |
2699 | ||
9079b1eb DH |
2700 | error = mnt_want_write(oldnd.path.mnt); |
2701 | if (error) | |
2702 | goto exit5; | |
be6d3e56 KT |
2703 | error = security_path_rename(&oldnd.path, old_dentry, |
2704 | &newnd.path, new_dentry); | |
2705 | if (error) | |
2706 | goto exit6; | |
1da177e4 LT |
2707 | error = vfs_rename(old_dir->d_inode, old_dentry, |
2708 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 2709 | exit6: |
9079b1eb | 2710 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
2711 | exit5: |
2712 | dput(new_dentry); | |
2713 | exit4: | |
2714 | dput(old_dentry); | |
2715 | exit3: | |
2716 | unlock_rename(new_dir, old_dir); | |
2717 | exit2: | |
1d957f9b | 2718 | path_put(&newnd.path); |
2ad94ae6 | 2719 | putname(to); |
1da177e4 | 2720 | exit1: |
1d957f9b | 2721 | path_put(&oldnd.path); |
1da177e4 | 2722 | putname(from); |
2ad94ae6 | 2723 | exit: |
1da177e4 LT |
2724 | return error; |
2725 | } | |
2726 | ||
a26eab24 | 2727 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2728 | { |
2729 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
2730 | } | |
2731 | ||
1da177e4 LT |
2732 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
2733 | { | |
2734 | int len; | |
2735 | ||
2736 | len = PTR_ERR(link); | |
2737 | if (IS_ERR(link)) | |
2738 | goto out; | |
2739 | ||
2740 | len = strlen(link); | |
2741 | if (len > (unsigned) buflen) | |
2742 | len = buflen; | |
2743 | if (copy_to_user(buffer, link, len)) | |
2744 | len = -EFAULT; | |
2745 | out: | |
2746 | return len; | |
2747 | } | |
2748 | ||
2749 | /* | |
2750 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
2751 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
2752 | * using) it for any given inode is up to filesystem. | |
2753 | */ | |
2754 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2755 | { | |
2756 | struct nameidata nd; | |
cc314eef | 2757 | void *cookie; |
694a1764 | 2758 | int res; |
cc314eef | 2759 | |
1da177e4 | 2760 | nd.depth = 0; |
cc314eef | 2761 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
2762 | if (IS_ERR(cookie)) |
2763 | return PTR_ERR(cookie); | |
2764 | ||
2765 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
2766 | if (dentry->d_inode->i_op->put_link) | |
2767 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
2768 | return res; | |
1da177e4 LT |
2769 | } |
2770 | ||
2771 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
2772 | { | |
2773 | return __vfs_follow_link(nd, link); | |
2774 | } | |
2775 | ||
2776 | /* get the link contents into pagecache */ | |
2777 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
2778 | { | |
ebd09abb DG |
2779 | char *kaddr; |
2780 | struct page *page; | |
1da177e4 | 2781 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 2782 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 2783 | if (IS_ERR(page)) |
6fe6900e | 2784 | return (char*)page; |
1da177e4 | 2785 | *ppage = page; |
ebd09abb DG |
2786 | kaddr = kmap(page); |
2787 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
2788 | return kaddr; | |
1da177e4 LT |
2789 | } |
2790 | ||
2791 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
2792 | { | |
2793 | struct page *page = NULL; | |
2794 | char *s = page_getlink(dentry, &page); | |
2795 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
2796 | if (page) { | |
2797 | kunmap(page); | |
2798 | page_cache_release(page); | |
2799 | } | |
2800 | return res; | |
2801 | } | |
2802 | ||
cc314eef | 2803 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 2804 | { |
cc314eef | 2805 | struct page *page = NULL; |
1da177e4 | 2806 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 2807 | return page; |
1da177e4 LT |
2808 | } |
2809 | ||
cc314eef | 2810 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 2811 | { |
cc314eef LT |
2812 | struct page *page = cookie; |
2813 | ||
2814 | if (page) { | |
1da177e4 LT |
2815 | kunmap(page); |
2816 | page_cache_release(page); | |
1da177e4 LT |
2817 | } |
2818 | } | |
2819 | ||
54566b2c NP |
2820 | /* |
2821 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
2822 | */ | |
2823 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
2824 | { |
2825 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 2826 | struct page *page; |
afddba49 | 2827 | void *fsdata; |
beb497ab | 2828 | int err; |
1da177e4 | 2829 | char *kaddr; |
54566b2c NP |
2830 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
2831 | if (nofs) | |
2832 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 2833 | |
7e53cac4 | 2834 | retry: |
afddba49 | 2835 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 2836 | flags, &page, &fsdata); |
1da177e4 | 2837 | if (err) |
afddba49 NP |
2838 | goto fail; |
2839 | ||
1da177e4 LT |
2840 | kaddr = kmap_atomic(page, KM_USER0); |
2841 | memcpy(kaddr, symname, len-1); | |
2842 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
2843 | |
2844 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
2845 | page, fsdata); | |
1da177e4 LT |
2846 | if (err < 0) |
2847 | goto fail; | |
afddba49 NP |
2848 | if (err < len-1) |
2849 | goto retry; | |
2850 | ||
1da177e4 LT |
2851 | mark_inode_dirty(inode); |
2852 | return 0; | |
1da177e4 LT |
2853 | fail: |
2854 | return err; | |
2855 | } | |
2856 | ||
0adb25d2 KK |
2857 | int page_symlink(struct inode *inode, const char *symname, int len) |
2858 | { | |
2859 | return __page_symlink(inode, symname, len, | |
54566b2c | 2860 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
2861 | } |
2862 | ||
92e1d5be | 2863 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
2864 | .readlink = generic_readlink, |
2865 | .follow_link = page_follow_link_light, | |
2866 | .put_link = page_put_link, | |
2867 | }; | |
2868 | ||
2d8f3038 | 2869 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
2870 | EXPORT_SYMBOL(follow_down); |
2871 | EXPORT_SYMBOL(follow_up); | |
2872 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
2873 | EXPORT_SYMBOL(getname); | |
2874 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
2875 | EXPORT_SYMBOL(lookup_one_len); |
2876 | EXPORT_SYMBOL(page_follow_link_light); | |
2877 | EXPORT_SYMBOL(page_put_link); | |
2878 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 2879 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
2880 | EXPORT_SYMBOL(page_symlink); |
2881 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
2882 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 2883 | EXPORT_SYMBOL(kern_path); |
16f18200 | 2884 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 2885 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 2886 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
2887 | EXPORT_SYMBOL(unlock_rename); |
2888 | EXPORT_SYMBOL(vfs_create); | |
2889 | EXPORT_SYMBOL(vfs_follow_link); | |
2890 | EXPORT_SYMBOL(vfs_link); | |
2891 | EXPORT_SYMBOL(vfs_mkdir); | |
2892 | EXPORT_SYMBOL(vfs_mknod); | |
2893 | EXPORT_SYMBOL(generic_permission); | |
2894 | EXPORT_SYMBOL(vfs_readlink); | |
2895 | EXPORT_SYMBOL(vfs_rename); | |
2896 | EXPORT_SYMBOL(vfs_rmdir); | |
2897 | EXPORT_SYMBOL(vfs_symlink); | |
2898 | EXPORT_SYMBOL(vfs_unlink); | |
2899 | EXPORT_SYMBOL(dentry_unhash); | |
2900 | EXPORT_SYMBOL(generic_readlink); |