blob: 241cff423653692ec067afa71ccf9c8ca4e86b05 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
Robert Love0eeca282005-07-12 17:06:03 -040024#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/personality.h>
26#include <linux/security.h>
27#include <linux/syscalls.h>
28#include <linux/mount.h>
29#include <linux/audit.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080030#include <linux/capability.h>
Trond Myklebust834f2a42005-10-18 14:20:16 -070031#include <linux/file.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080032#include <linux/fcntl.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/namei.h>
34#include <asm/uaccess.h>
35
36#define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE])
37
38/* [Feb-1997 T. Schoebel-Theuer]
39 * Fundamental changes in the pathname lookup mechanisms (namei)
40 * were necessary because of omirr. The reason is that omirr needs
41 * to know the _real_ pathname, not the user-supplied one, in case
42 * of symlinks (and also when transname replacements occur).
43 *
44 * The new code replaces the old recursive symlink resolution with
45 * an iterative one (in case of non-nested symlink chains). It does
46 * this with calls to <fs>_follow_link().
47 * As a side effect, dir_namei(), _namei() and follow_link() are now
48 * replaced with a single function lookup_dentry() that can handle all
49 * the special cases of the former code.
50 *
51 * With the new dcache, the pathname is stored at each inode, at least as
52 * long as the refcount of the inode is positive. As a side effect, the
53 * size of the dcache depends on the inode cache and thus is dynamic.
54 *
55 * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink
56 * resolution to correspond with current state of the code.
57 *
58 * Note that the symlink resolution is not *completely* iterative.
59 * There is still a significant amount of tail- and mid- recursion in
60 * the algorithm. Also, note that <fs>_readlink() is not used in
61 * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink()
62 * may return different results than <fs>_follow_link(). Many virtual
63 * filesystems (including /proc) exhibit this behavior.
64 */
65
66/* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation:
67 * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL
68 * and the name already exists in form of a symlink, try to create the new
69 * name indicated by the symlink. The old code always complained that the
70 * name already exists, due to not following the symlink even if its target
71 * is nonexistent. The new semantics affects also mknod() and link() when
72 * the name is a symlink pointing to a non-existant name.
73 *
74 * I don't know which semantics is the right one, since I have no access
75 * to standards. But I found by trial that HP-UX 9.0 has the full "new"
76 * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the
77 * "old" one. Personally, I think the new semantics is much more logical.
78 * Note that "ln old new" where "new" is a symlink pointing to a non-existing
79 * file does succeed in both HP-UX and SunOs, but not in Solaris
80 * and in the old Linux semantics.
81 */
82
83/* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink
84 * semantics. See the comments in "open_namei" and "do_link" below.
85 *
86 * [10-Sep-98 Alan Modra] Another symlink change.
87 */
88
89/* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks:
90 * inside the path - always follow.
91 * in the last component in creation/removal/renaming - never follow.
92 * if LOOKUP_FOLLOW passed - follow.
93 * if the pathname has trailing slashes - follow.
94 * otherwise - don't follow.
95 * (applied in that order).
96 *
97 * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT
98 * restored for 2.4. This is the last surviving part of old 4.2BSD bug.
99 * During the 2.4 we need to fix the userland stuff depending on it -
100 * hopefully we will be able to get rid of that wart in 2.5. So far only
101 * XEmacs seems to be relying on it...
102 */
103/*
104 * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland)
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800105 * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 * any extra contention...
107 */
108
Josef 'Jeff' Sipekc4a78082007-07-19 01:48:22 -0700109static int fastcall link_path_walk(const char *name, struct nameidata *nd);
110
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111/* 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 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800118static int do_getname(const char __user *filename, char *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119{
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
140char * 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
160void putname(const char *name)
161{
Al Viro5ac3a9c2006-07-16 06:38:45 -0400162 if (unlikely(!audit_dummy_context()))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 audit_putname(name);
164 else
165 __putname(name);
166}
167EXPORT_SYMBOL(putname);
168#endif
169
170
171/**
172 * generic_permission - check for access rights on a Posix-like filesystem
173 * @inode: inode to check access rights for
174 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
175 * @check_acl: optional callback to check for Posix ACLs
176 *
177 * Used to check for read/write/execute permissions on a file.
178 * We use "fsuid" for this, letting us set arbitrary permissions
179 * for filesystem access without changing the "normal" uids which
180 * are used for other things..
181 */
182int generic_permission(struct inode *inode, int mask,
183 int (*check_acl)(struct inode *inode, int mask))
184{
185 umode_t mode = inode->i_mode;
186
187 if (current->fsuid == inode->i_uid)
188 mode >>= 6;
189 else {
190 if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) {
191 int error = check_acl(inode, mask);
192 if (error == -EACCES)
193 goto check_capabilities;
194 else if (error != -EAGAIN)
195 return error;
196 }
197
198 if (in_group_p(inode->i_gid))
199 mode >>= 3;
200 }
201
202 /*
203 * If the DACs are ok we don't need any capability check.
204 */
205 if (((mode & mask & (MAY_READ|MAY_WRITE|MAY_EXEC)) == mask))
206 return 0;
207
208 check_capabilities:
209 /*
210 * Read/write DACs are always overridable.
211 * Executable DACs are overridable if at least one exec bit is set.
212 */
213 if (!(mask & MAY_EXEC) ||
214 (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode))
215 if (capable(CAP_DAC_OVERRIDE))
216 return 0;
217
218 /*
219 * Searching includes executable on directories, else just read.
220 */
221 if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE)))
222 if (capable(CAP_DAC_READ_SEARCH))
223 return 0;
224
225 return -EACCES;
226}
227
228int permission(struct inode *inode, int mask, struct nameidata *nd)
229{
230 int retval, submask;
Dave Hansenc7eb2662007-10-16 23:31:14 -0700231 struct vfsmount *mnt = NULL;
232
233 if (nd)
234 mnt = nd->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235
236 if (mask & MAY_WRITE) {
Miklos Szeredi22590e42007-10-16 23:27:08 -0700237 umode_t mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239 /*
240 * Nobody gets write access to a read-only fs.
241 */
242 if (IS_RDONLY(inode) &&
243 (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
244 return -EROFS;
245
246 /*
247 * Nobody gets write access to an immutable file.
248 */
249 if (IS_IMMUTABLE(inode))
250 return -EACCES;
251 }
252
Miklos Szeredi22590e42007-10-16 23:27:08 -0700253 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) {
254 /*
255 * MAY_EXEC on regular files is denied if the fs is mounted
256 * with the "noexec" flag.
257 */
Dave Hansenc7eb2662007-10-16 23:31:14 -0700258 if (mnt && (mnt->mnt_flags & MNT_NOEXEC))
Miklos Szeredi22590e42007-10-16 23:27:08 -0700259 return -EACCES;
260 }
Trond Myklebusta343bb72006-08-22 20:06:03 -0400261
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 /* Ordinary permission routines do not understand MAY_APPEND. */
263 submask = mask & ~MAY_APPEND;
Miklos Szeredi22590e42007-10-16 23:27:08 -0700264 if (inode->i_op && inode->i_op->permission) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 retval = inode->i_op->permission(inode, submask, nd);
Miklos Szeredi22590e42007-10-16 23:27:08 -0700266 if (!retval) {
267 /*
268 * Exec permission on a regular file is denied if none
269 * of the execute bits are set.
270 *
271 * This check should be done by the ->permission()
272 * method.
273 */
274 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode) &&
275 !(inode->i_mode & S_IXUGO))
276 return -EACCES;
277 }
278 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 retval = generic_permission(inode, submask, NULL);
Miklos Szeredi22590e42007-10-16 23:27:08 -0700280 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 if (retval)
282 return retval;
283
284 return security_inode_permission(inode, mask, nd);
285}
286
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800287/**
288 * vfs_permission - check for access rights to a given path
289 * @nd: lookup result that describes the path
290 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
291 *
292 * Used to check for read/write/execute permissions on a path.
293 * We use "fsuid" for this, letting us set arbitrary permissions
294 * for filesystem access without changing the "normal" uids which
295 * are used for other things.
296 */
297int vfs_permission(struct nameidata *nd, int mask)
298{
299 return permission(nd->dentry->d_inode, mask, nd);
300}
301
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800302/**
303 * file_permission - check for additional access rights to a given file
304 * @file: file to check access rights for
305 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
306 *
307 * Used to check for read/write/execute permissions on an already opened
308 * file.
309 *
310 * Note:
311 * Do not use this function in new code. All access checks should
312 * be done using vfs_permission().
313 */
314int file_permission(struct file *file, int mask)
315{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800316 return permission(file->f_path.dentry->d_inode, mask, NULL);
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800317}
318
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319/*
320 * get_write_access() gets write permission for a file.
321 * put_write_access() releases this write permission.
322 * This is used for regular files.
323 * We cannot support write (and maybe mmap read-write shared) accesses and
324 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
325 * can have the following values:
326 * 0: no writers, no VM_DENYWRITE mappings
327 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
328 * > 0: (i_writecount) users are writing to the file.
329 *
330 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
331 * except for the cases where we don't hold i_writecount yet. Then we need to
332 * use {get,deny}_write_access() - these functions check the sign and refuse
333 * to do the change if sign is wrong. Exclusion between them is provided by
334 * the inode->i_lock spinlock.
335 */
336
337int get_write_access(struct inode * inode)
338{
339 spin_lock(&inode->i_lock);
340 if (atomic_read(&inode->i_writecount) < 0) {
341 spin_unlock(&inode->i_lock);
342 return -ETXTBSY;
343 }
344 atomic_inc(&inode->i_writecount);
345 spin_unlock(&inode->i_lock);
346
347 return 0;
348}
349
350int deny_write_access(struct file * file)
351{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800352 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353
354 spin_lock(&inode->i_lock);
355 if (atomic_read(&inode->i_writecount) > 0) {
356 spin_unlock(&inode->i_lock);
357 return -ETXTBSY;
358 }
359 atomic_dec(&inode->i_writecount);
360 spin_unlock(&inode->i_lock);
361
362 return 0;
363}
364
365void path_release(struct nameidata *nd)
366{
367 dput(nd->dentry);
368 mntput(nd->mnt);
369}
370
371/*
372 * umount() mustn't call path_release()/mntput() as that would clear
373 * mnt_expiry_mark
374 */
375void path_release_on_umount(struct nameidata *nd)
376{
377 dput(nd->dentry);
Miklos Szeredi751c4042005-07-07 17:57:30 -0700378 mntput_no_expire(nd->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379}
380
Trond Myklebust834f2a42005-10-18 14:20:16 -0700381/**
382 * release_open_intent - free up open intent resources
383 * @nd: pointer to nameidata
384 */
385void release_open_intent(struct nameidata *nd)
386{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800387 if (nd->intent.open.file->f_path.dentry == NULL)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700388 put_filp(nd->intent.open.file);
389 else
390 fput(nd->intent.open.file);
391}
392
Ian Kentbcdc5e02006-09-27 01:50:44 -0700393static inline struct dentry *
394do_revalidate(struct dentry *dentry, struct nameidata *nd)
395{
396 int status = dentry->d_op->d_revalidate(dentry, nd);
397 if (unlikely(status <= 0)) {
398 /*
399 * The dentry failed validation.
400 * If d_revalidate returned 0 attempt to invalidate
401 * the dentry otherwise d_revalidate is asking us
402 * to return a fail status.
403 */
404 if (!status) {
405 if (!d_invalidate(dentry)) {
406 dput(dentry);
407 dentry = NULL;
408 }
409 } else {
410 dput(dentry);
411 dentry = ERR_PTR(status);
412 }
413 }
414 return dentry;
415}
416
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417/*
418 * Internal lookup() using the new generic dcache.
419 * SMP-safe
420 */
421static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd)
422{
423 struct dentry * dentry = __d_lookup(parent, name);
424
425 /* lockess __d_lookup may fail due to concurrent d_move()
426 * in some unrelated directory, so try with d_lookup
427 */
428 if (!dentry)
429 dentry = d_lookup(parent, name);
430
Ian Kentbcdc5e02006-09-27 01:50:44 -0700431 if (dentry && dentry->d_op && dentry->d_op->d_revalidate)
432 dentry = do_revalidate(dentry, nd);
433
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 return dentry;
435}
436
437/*
438 * Short-cut version of permission(), for calling by
439 * path_walk(), when dcache lock is held. Combines parts
440 * of permission() and generic_permission(), and tests ONLY for
441 * MAY_EXEC permission.
442 *
443 * If appropriate, check DAC only. If not appropriate, or
444 * short-cut DAC fails, then call permission() to do more
445 * complete permission check.
446 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800447static int exec_permission_lite(struct inode *inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 struct nameidata *nd)
449{
450 umode_t mode = inode->i_mode;
451
452 if (inode->i_op && inode->i_op->permission)
453 return -EAGAIN;
454
455 if (current->fsuid == inode->i_uid)
456 mode >>= 6;
457 else if (in_group_p(inode->i_gid))
458 mode >>= 3;
459
460 if (mode & MAY_EXEC)
461 goto ok;
462
463 if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE))
464 goto ok;
465
466 if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE))
467 goto ok;
468
469 if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH))
470 goto ok;
471
472 return -EACCES;
473ok:
474 return security_inode_permission(inode, MAY_EXEC, nd);
475}
476
477/*
478 * This is called when everything else fails, and we actually have
479 * to go to the low-level filesystem to find out what we should do..
480 *
481 * We get the directory semaphore, and after getting that we also
482 * make sure that nobody added the entry to the dcache in the meantime..
483 * SMP-safe
484 */
485static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd)
486{
487 struct dentry * result;
488 struct inode *dir = parent->d_inode;
489
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800490 mutex_lock(&dir->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 /*
492 * First re-do the cached lookup just in case it was created
493 * while we waited for the directory semaphore..
494 *
495 * FIXME! This could use version numbering or similar to
496 * avoid unnecessary cache lookups.
497 *
498 * The "dcache_lock" is purely to protect the RCU list walker
499 * from concurrent renames at this point (we mustn't get false
500 * negatives from the RCU list walk here, unlike the optimistic
501 * fast walk).
502 *
503 * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup
504 */
505 result = d_lookup(parent, name);
506 if (!result) {
507 struct dentry * dentry = d_alloc(parent, name);
508 result = ERR_PTR(-ENOMEM);
509 if (dentry) {
510 result = dir->i_op->lookup(dir, dentry, nd);
511 if (result)
512 dput(dentry);
513 else
514 result = dentry;
515 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800516 mutex_unlock(&dir->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 return result;
518 }
519
520 /*
521 * Uhhuh! Nasty case: the cache was re-populated while
522 * we waited on the semaphore. Need to revalidate.
523 */
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800524 mutex_unlock(&dir->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (result->d_op && result->d_op->d_revalidate) {
Ian Kentbcdc5e02006-09-27 01:50:44 -0700526 result = do_revalidate(result, nd);
527 if (!result)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 }
530 return result;
531}
532
533static int __emul_lookup_dentry(const char *, struct nameidata *);
534
535/* SMP-safe */
Arjan van de Venf1662352006-01-14 13:21:31 -0800536static __always_inline int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537walk_init_root(const char *name, struct nameidata *nd)
538{
Andreas Mohre518ddb2006-09-29 02:01:22 -0700539 struct fs_struct *fs = current->fs;
540
541 read_lock(&fs->lock);
542 if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) {
543 nd->mnt = mntget(fs->altrootmnt);
544 nd->dentry = dget(fs->altroot);
545 read_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 if (__emul_lookup_dentry(name,nd))
547 return 0;
Andreas Mohre518ddb2006-09-29 02:01:22 -0700548 read_lock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 }
Andreas Mohre518ddb2006-09-29 02:01:22 -0700550 nd->mnt = mntget(fs->rootmnt);
551 nd->dentry = dget(fs->root);
552 read_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 return 1;
554}
555
Arjan van de Venf1662352006-01-14 13:21:31 -0800556static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557{
558 int res = 0;
559 char *name;
560 if (IS_ERR(link))
561 goto fail;
562
563 if (*link == '/') {
564 path_release(nd);
565 if (!walk_init_root(link, nd))
566 /* weird __emul_prefix() stuff did it */
567 goto out;
568 }
569 res = link_path_walk(link, nd);
570out:
571 if (nd->depth || res || nd->last_type!=LAST_NORM)
572 return res;
573 /*
574 * If it is an iterative symlinks resolution in open_namei() we
575 * have to copy the last component. And all that crap because of
576 * bloody create() on broken symlinks. Furrfu...
577 */
578 name = __getname();
579 if (unlikely(!name)) {
580 path_release(nd);
581 return -ENOMEM;
582 }
583 strcpy(name, nd->last.name);
584 nd->last.name = name;
585 return 0;
586fail:
587 path_release(nd);
588 return PTR_ERR(link);
589}
590
Miklos Szeredi09dd17d2005-09-06 15:18:21 -0700591static inline void dput_path(struct path *path, struct nameidata *nd)
592{
593 dput(path->dentry);
594 if (path->mnt != nd->mnt)
595 mntput(path->mnt);
596}
597
598static inline void path_to_nameidata(struct path *path, struct nameidata *nd)
599{
600 dput(nd->dentry);
601 if (nd->mnt != path->mnt)
602 mntput(nd->mnt);
603 nd->mnt = path->mnt;
604 nd->dentry = path->dentry;
605}
606
Ian Kent051d3812006-03-27 01:14:53 -0800607static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd)
608{
609 int error;
610 void *cookie;
611 struct dentry *dentry = path->dentry;
612
613 touch_atime(path->mnt, dentry);
614 nd_set_link(nd, NULL);
615
616 if (path->mnt != nd->mnt) {
617 path_to_nameidata(path, nd);
618 dget(dentry);
619 }
620 mntget(path->mnt);
621 cookie = dentry->d_inode->i_op->follow_link(dentry, nd);
622 error = PTR_ERR(cookie);
623 if (!IS_ERR(cookie)) {
624 char *s = nd_get_link(nd);
625 error = 0;
626 if (s)
627 error = __vfs_follow_link(nd, s);
628 if (dentry->d_inode->i_op->put_link)
629 dentry->d_inode->i_op->put_link(dentry, nd, cookie);
630 }
631 dput(dentry);
632 mntput(path->mnt);
633
634 return error;
635}
636
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637/*
638 * This limits recursive symlink follows to 8, while
639 * limiting consecutive symlinks to 40.
640 *
641 * Without that kind of total limit, nasty chains of consecutive
642 * symlinks can cause almost arbitrarily long lookups.
643 */
Al Viro90ebe562005-06-06 13:35:58 -0700644static inline int do_follow_link(struct path *path, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645{
646 int err = -ELOOP;
647 if (current->link_count >= MAX_NESTED_LINKS)
648 goto loop;
649 if (current->total_link_count >= 40)
650 goto loop;
651 BUG_ON(nd->depth >= MAX_NESTED_LINKS);
652 cond_resched();
Al Viro90ebe562005-06-06 13:35:58 -0700653 err = security_inode_follow_link(path->dentry, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 if (err)
655 goto loop;
656 current->link_count++;
657 current->total_link_count++;
658 nd->depth++;
Al Virocd4e91d2005-06-06 13:36:03 -0700659 err = __do_follow_link(path, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 current->link_count--;
661 nd->depth--;
662 return err;
663loop:
Miklos Szeredi09dd17d2005-09-06 15:18:21 -0700664 dput_path(path, nd);
Al Viro839d9f92005-06-06 13:36:02 -0700665 path_release(nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 return err;
667}
668
669int follow_up(struct vfsmount **mnt, struct dentry **dentry)
670{
671 struct vfsmount *parent;
672 struct dentry *mountpoint;
673 spin_lock(&vfsmount_lock);
674 parent=(*mnt)->mnt_parent;
675 if (parent == *mnt) {
676 spin_unlock(&vfsmount_lock);
677 return 0;
678 }
679 mntget(parent);
680 mountpoint=dget((*mnt)->mnt_mountpoint);
681 spin_unlock(&vfsmount_lock);
682 dput(*dentry);
683 *dentry = mountpoint;
684 mntput(*mnt);
685 *mnt = parent;
686 return 1;
687}
688
689/* no need for dcache_lock, as serialization is taken care in
690 * namespace.c
691 */
Al Viro463ffb22005-06-06 13:36:05 -0700692static int __follow_mount(struct path *path)
693{
694 int res = 0;
695 while (d_mountpoint(path->dentry)) {
696 struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry);
697 if (!mounted)
698 break;
699 dput(path->dentry);
700 if (res)
701 mntput(path->mnt);
702 path->mnt = mounted;
703 path->dentry = dget(mounted->mnt_root);
704 res = 1;
705 }
706 return res;
707}
708
Al Viro58c465e2005-06-06 13:36:13 -0700709static void follow_mount(struct vfsmount **mnt, struct dentry **dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 while (d_mountpoint(*dentry)) {
712 struct vfsmount *mounted = lookup_mnt(*mnt, *dentry);
713 if (!mounted)
714 break;
Al Viro58c465e2005-06-06 13:36:13 -0700715 dput(*dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 mntput(*mnt);
717 *mnt = mounted;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 *dentry = dget(mounted->mnt_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720}
721
722/* no need for dcache_lock, as serialization is taken care in
723 * namespace.c
724 */
Al Viroe13b2102005-06-06 13:36:06 -0700725int follow_down(struct vfsmount **mnt, struct dentry **dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726{
727 struct vfsmount *mounted;
728
729 mounted = lookup_mnt(*mnt, *dentry);
730 if (mounted) {
Al Viroe13b2102005-06-06 13:36:06 -0700731 dput(*dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 mntput(*mnt);
733 *mnt = mounted;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 *dentry = dget(mounted->mnt_root);
735 return 1;
736 }
737 return 0;
738}
739
Arjan van de Venf1662352006-01-14 13:21:31 -0800740static __always_inline void follow_dotdot(struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741{
Andreas Mohre518ddb2006-09-29 02:01:22 -0700742 struct fs_struct *fs = current->fs;
743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 while(1) {
745 struct vfsmount *parent;
Al Viro58c465e2005-06-06 13:36:13 -0700746 struct dentry *old = nd->dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
Andreas Mohre518ddb2006-09-29 02:01:22 -0700748 read_lock(&fs->lock);
749 if (nd->dentry == fs->root &&
750 nd->mnt == fs->rootmnt) {
751 read_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 break;
753 }
Andreas Mohre518ddb2006-09-29 02:01:22 -0700754 read_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 spin_lock(&dcache_lock);
Al Viro58c465e2005-06-06 13:36:13 -0700756 if (nd->dentry != nd->mnt->mnt_root) {
757 nd->dentry = dget(nd->dentry->d_parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 spin_unlock(&dcache_lock);
759 dput(old);
760 break;
761 }
762 spin_unlock(&dcache_lock);
763 spin_lock(&vfsmount_lock);
Al Viro58c465e2005-06-06 13:36:13 -0700764 parent = nd->mnt->mnt_parent;
765 if (parent == nd->mnt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 spin_unlock(&vfsmount_lock);
767 break;
768 }
769 mntget(parent);
Al Viro58c465e2005-06-06 13:36:13 -0700770 nd->dentry = dget(nd->mnt->mnt_mountpoint);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 spin_unlock(&vfsmount_lock);
772 dput(old);
Al Viro58c465e2005-06-06 13:36:13 -0700773 mntput(nd->mnt);
774 nd->mnt = parent;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
Al Viro58c465e2005-06-06 13:36:13 -0700776 follow_mount(&nd->mnt, &nd->dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777}
778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779/*
780 * It's more convoluted than I'd like it to be, but... it's still fairly
781 * small and for now I'd prefer to have fast path as straight as possible.
782 * It _is_ time-critical.
783 */
784static int do_lookup(struct nameidata *nd, struct qstr *name,
785 struct path *path)
786{
787 struct vfsmount *mnt = nd->mnt;
788 struct dentry *dentry = __d_lookup(nd->dentry, name);
789
790 if (!dentry)
791 goto need_lookup;
792 if (dentry->d_op && dentry->d_op->d_revalidate)
793 goto need_revalidate;
794done:
795 path->mnt = mnt;
796 path->dentry = dentry;
Al Viro634ee702005-06-06 13:36:13 -0700797 __follow_mount(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 return 0;
799
800need_lookup:
801 dentry = real_lookup(nd->dentry, name, nd);
802 if (IS_ERR(dentry))
803 goto fail;
804 goto done;
805
806need_revalidate:
Ian Kentbcdc5e02006-09-27 01:50:44 -0700807 dentry = do_revalidate(dentry, nd);
808 if (!dentry)
809 goto need_lookup;
810 if (IS_ERR(dentry))
811 goto fail;
812 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
814fail:
815 return PTR_ERR(dentry);
816}
817
818/*
819 * Name resolution.
Prasanna Medaea3834d2005-04-29 16:00:17 +0100820 * This is the basic name resolution function, turning a pathname into
821 * the final dentry. We expect 'base' to be positive and a directory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 *
Prasanna Medaea3834d2005-04-29 16:00:17 +0100823 * Returns 0 and nd will have valid dentry and mnt on success.
824 * Returns error and drops reference to input namei data on failure.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 */
826static fastcall int __link_path_walk(const char * name, struct nameidata *nd)
827{
828 struct path next;
829 struct inode *inode;
830 int err;
831 unsigned int lookup_flags = nd->flags;
832
833 while (*name=='/')
834 name++;
835 if (!*name)
836 goto return_reval;
837
838 inode = nd->dentry->d_inode;
839 if (nd->depth)
Trond Myklebustf55eab82006-02-04 23:28:01 -0800840 lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
842 /* At this point we know we have a real path component. */
843 for(;;) {
844 unsigned long hash;
845 struct qstr this;
846 unsigned int c;
847
Trond Myklebustcdce5d62005-10-18 14:20:18 -0700848 nd->flags |= LOOKUP_CONTINUE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 err = exec_permission_lite(inode, nd);
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800850 if (err == -EAGAIN)
851 err = vfs_permission(nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 if (err)
853 break;
854
855 this.name = name;
856 c = *(const unsigned char *)name;
857
858 hash = init_name_hash();
859 do {
860 name++;
861 hash = partial_name_hash(c, hash);
862 c = *(const unsigned char *)name;
863 } while (c && (c != '/'));
864 this.len = name - (const char *) this.name;
865 this.hash = end_name_hash(hash);
866
867 /* remove trailing slashes? */
868 if (!c)
869 goto last_component;
870 while (*++name == '/');
871 if (!*name)
872 goto last_with_slashes;
873
874 /*
875 * "." and ".." are special - ".." especially so because it has
876 * to be able to know about the current root directory and
877 * parent relationships.
878 */
879 if (this.name[0] == '.') switch (this.len) {
880 default:
881 break;
882 case 2:
883 if (this.name[1] != '.')
884 break;
Al Viro58c465e2005-06-06 13:36:13 -0700885 follow_dotdot(nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 inode = nd->dentry->d_inode;
887 /* fallthrough */
888 case 1:
889 continue;
890 }
891 /*
892 * See if the low-level filesystem might want
893 * to use its own hash..
894 */
895 if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
896 err = nd->dentry->d_op->d_hash(nd->dentry, &this);
897 if (err < 0)
898 break;
899 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 /* This does the actual lookups.. */
901 err = do_lookup(nd, &this, &next);
902 if (err)
903 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
905 err = -ENOENT;
906 inode = next.dentry->d_inode;
907 if (!inode)
908 goto out_dput;
909 err = -ENOTDIR;
910 if (!inode->i_op)
911 goto out_dput;
912
913 if (inode->i_op->follow_link) {
Al Viro90ebe562005-06-06 13:35:58 -0700914 err = do_follow_link(&next, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 if (err)
916 goto return_err;
917 err = -ENOENT;
918 inode = nd->dentry->d_inode;
919 if (!inode)
920 break;
921 err = -ENOTDIR;
922 if (!inode->i_op)
923 break;
Miklos Szeredi09dd17d2005-09-06 15:18:21 -0700924 } else
925 path_to_nameidata(&next, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 err = -ENOTDIR;
927 if (!inode->i_op->lookup)
928 break;
929 continue;
930 /* here ends the main loop */
931
932last_with_slashes:
933 lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
934last_component:
Trond Myklebustf55eab82006-02-04 23:28:01 -0800935 /* Clear LOOKUP_CONTINUE iff it was previously unset */
936 nd->flags &= lookup_flags | ~LOOKUP_CONTINUE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 if (lookup_flags & LOOKUP_PARENT)
938 goto lookup_parent;
939 if (this.name[0] == '.') switch (this.len) {
940 default:
941 break;
942 case 2:
943 if (this.name[1] != '.')
944 break;
Al Viro58c465e2005-06-06 13:36:13 -0700945 follow_dotdot(nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 inode = nd->dentry->d_inode;
947 /* fallthrough */
948 case 1:
949 goto return_reval;
950 }
951 if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
952 err = nd->dentry->d_op->d_hash(nd->dentry, &this);
953 if (err < 0)
954 break;
955 }
956 err = do_lookup(nd, &this, &next);
957 if (err)
958 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 inode = next.dentry->d_inode;
960 if ((lookup_flags & LOOKUP_FOLLOW)
961 && inode && inode->i_op && inode->i_op->follow_link) {
Al Viro90ebe562005-06-06 13:35:58 -0700962 err = do_follow_link(&next, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 if (err)
964 goto return_err;
965 inode = nd->dentry->d_inode;
Miklos Szeredi09dd17d2005-09-06 15:18:21 -0700966 } else
967 path_to_nameidata(&next, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 err = -ENOENT;
969 if (!inode)
970 break;
971 if (lookup_flags & LOOKUP_DIRECTORY) {
972 err = -ENOTDIR;
973 if (!inode->i_op || !inode->i_op->lookup)
974 break;
975 }
976 goto return_base;
977lookup_parent:
978 nd->last = this;
979 nd->last_type = LAST_NORM;
980 if (this.name[0] != '.')
981 goto return_base;
982 if (this.len == 1)
983 nd->last_type = LAST_DOT;
984 else if (this.len == 2 && this.name[1] == '.')
985 nd->last_type = LAST_DOTDOT;
986 else
987 goto return_base;
988return_reval:
989 /*
990 * We bypassed the ordinary revalidation routines.
991 * We may need to check the cached dentry for staleness.
992 */
993 if (nd->dentry && nd->dentry->d_sb &&
994 (nd->dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) {
995 err = -ESTALE;
996 /* Note: we do not d_invalidate() */
997 if (!nd->dentry->d_op->d_revalidate(nd->dentry, nd))
998 break;
999 }
1000return_base:
1001 return 0;
1002out_dput:
Miklos Szeredi09dd17d2005-09-06 15:18:21 -07001003 dput_path(&next, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 break;
1005 }
1006 path_release(nd);
1007return_err:
1008 return err;
1009}
1010
1011/*
1012 * Wrapper to retry pathname resolution whenever the underlying
1013 * file system returns an ESTALE.
1014 *
1015 * Retry the whole path once, forcing real lookup requests
1016 * instead of relying on the dcache.
1017 */
Josef 'Jeff' Sipekc4a78082007-07-19 01:48:22 -07001018static int fastcall link_path_walk(const char *name, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019{
1020 struct nameidata save = *nd;
1021 int result;
1022
1023 /* make sure the stuff we saved doesn't go away */
1024 dget(save.dentry);
1025 mntget(save.mnt);
1026
1027 result = __link_path_walk(name, nd);
1028 if (result == -ESTALE) {
1029 *nd = save;
1030 dget(nd->dentry);
1031 mntget(nd->mnt);
1032 nd->flags |= LOOKUP_REVAL;
1033 result = __link_path_walk(name, nd);
1034 }
1035
1036 dput(save.dentry);
1037 mntput(save.mnt);
1038
1039 return result;
1040}
1041
Josef 'Jeff' Sipekf79c20f2007-07-19 01:48:22 -07001042static int fastcall path_walk(const char * name, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043{
1044 current->total_link_count = 0;
1045 return link_path_walk(name, nd);
1046}
1047
Prasanna Medaea3834d2005-04-29 16:00:17 +01001048/*
1049 * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if
1050 * everything is done. Returns 0 and drops input nd, if lookup failed;
1051 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052static int __emul_lookup_dentry(const char *name, struct nameidata *nd)
1053{
1054 if (path_walk(name, nd))
1055 return 0; /* something went wrong... */
1056
1057 if (!nd->dentry->d_inode || S_ISDIR(nd->dentry->d_inode->i_mode)) {
1058 struct dentry *old_dentry = nd->dentry;
1059 struct vfsmount *old_mnt = nd->mnt;
1060 struct qstr last = nd->last;
1061 int last_type = nd->last_type;
Andreas Mohre518ddb2006-09-29 02:01:22 -07001062 struct fs_struct *fs = current->fs;
1063
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 /*
Andreas Mohre518ddb2006-09-29 02:01:22 -07001065 * NAME was not found in alternate root or it's a directory.
1066 * Try to find it in the normal root:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 */
1068 nd->last_type = LAST_ROOT;
Andreas Mohre518ddb2006-09-29 02:01:22 -07001069 read_lock(&fs->lock);
1070 nd->mnt = mntget(fs->rootmnt);
1071 nd->dentry = dget(fs->root);
1072 read_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 if (path_walk(name, nd) == 0) {
1074 if (nd->dentry->d_inode) {
1075 dput(old_dentry);
1076 mntput(old_mnt);
1077 return 1;
1078 }
1079 path_release(nd);
1080 }
1081 nd->dentry = old_dentry;
1082 nd->mnt = old_mnt;
1083 nd->last = last;
1084 nd->last_type = last_type;
1085 }
1086 return 1;
1087}
1088
1089void set_fs_altroot(void)
1090{
1091 char *emul = __emul_prefix();
1092 struct nameidata nd;
1093 struct vfsmount *mnt = NULL, *oldmnt;
1094 struct dentry *dentry = NULL, *olddentry;
1095 int err;
Andreas Mohre518ddb2006-09-29 02:01:22 -07001096 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
1098 if (!emul)
1099 goto set_it;
1100 err = path_lookup(emul, LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_NOALT, &nd);
1101 if (!err) {
1102 mnt = nd.mnt;
1103 dentry = nd.dentry;
1104 }
1105set_it:
Andreas Mohre518ddb2006-09-29 02:01:22 -07001106 write_lock(&fs->lock);
1107 oldmnt = fs->altrootmnt;
1108 olddentry = fs->altroot;
1109 fs->altrootmnt = mnt;
1110 fs->altroot = dentry;
1111 write_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 if (olddentry) {
1113 dput(olddentry);
1114 mntput(oldmnt);
1115 }
1116}
1117
Prasanna Medaea3834d2005-04-29 16:00:17 +01001118/* Returns 0 and nd will be valid on success; Retuns error, otherwise. */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001119static int fastcall do_path_lookup(int dfd, const char *name,
1120 unsigned int flags, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121{
Prasanna Medaea3834d2005-04-29 16:00:17 +01001122 int retval = 0;
Ulrich Drepper170aa3d2006-02-04 23:28:02 -08001123 int fput_needed;
1124 struct file *file;
Andreas Mohre518ddb2006-09-29 02:01:22 -07001125 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
1127 nd->last_type = LAST_ROOT; /* if there are only slashes... */
1128 nd->flags = flags;
1129 nd->depth = 0;
1130
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 if (*name=='/') {
Andreas Mohre518ddb2006-09-29 02:01:22 -07001132 read_lock(&fs->lock);
1133 if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) {
1134 nd->mnt = mntget(fs->altrootmnt);
1135 nd->dentry = dget(fs->altroot);
1136 read_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 if (__emul_lookup_dentry(name,nd))
Prasanna Medaea3834d2005-04-29 16:00:17 +01001138 goto out; /* found in altroot */
Andreas Mohre518ddb2006-09-29 02:01:22 -07001139 read_lock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 }
Andreas Mohre518ddb2006-09-29 02:01:22 -07001141 nd->mnt = mntget(fs->rootmnt);
1142 nd->dentry = dget(fs->root);
1143 read_unlock(&fs->lock);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001144 } else if (dfd == AT_FDCWD) {
Andreas Mohre518ddb2006-09-29 02:01:22 -07001145 read_lock(&fs->lock);
1146 nd->mnt = mntget(fs->pwdmnt);
1147 nd->dentry = dget(fs->pwd);
1148 read_unlock(&fs->lock);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001149 } else {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001150 struct dentry *dentry;
1151
1152 file = fget_light(dfd, &fput_needed);
Ulrich Drepper170aa3d2006-02-04 23:28:02 -08001153 retval = -EBADF;
1154 if (!file)
Trond Myklebust6d09bb62006-06-04 02:51:37 -07001155 goto out_fail;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001156
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001157 dentry = file->f_path.dentry;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001158
Ulrich Drepper170aa3d2006-02-04 23:28:02 -08001159 retval = -ENOTDIR;
1160 if (!S_ISDIR(dentry->d_inode->i_mode))
Trond Myklebust6d09bb62006-06-04 02:51:37 -07001161 goto fput_fail;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001162
1163 retval = file_permission(file, MAY_EXEC);
Ulrich Drepper170aa3d2006-02-04 23:28:02 -08001164 if (retval)
Trond Myklebust6d09bb62006-06-04 02:51:37 -07001165 goto fput_fail;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001166
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001167 nd->mnt = mntget(file->f_path.mnt);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001168 nd->dentry = dget(dentry);
1169
1170 fput_light(file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 }
Josef 'Jeff' Sipek2dfdd262007-05-09 02:33:41 -07001172
1173 retval = path_walk(name, nd);
Prasanna Medaea3834d2005-04-29 16:00:17 +01001174out:
Josef 'Jeff' Sipek62ce39c2007-05-09 02:33:41 -07001175 if (unlikely(!retval && !audit_dummy_context() && nd->dentry &&
Suzuki3bc84142006-02-07 12:58:36 -08001176 nd->dentry->d_inode))
Al Viro5a190ae2007-06-07 12:19:32 -04001177 audit_inode(name, nd->dentry);
Trond Myklebust6d09bb62006-06-04 02:51:37 -07001178out_fail:
Ulrich Drepper170aa3d2006-02-04 23:28:02 -08001179 return retval;
1180
Trond Myklebust6d09bb62006-06-04 02:51:37 -07001181fput_fail:
Ulrich Drepper170aa3d2006-02-04 23:28:02 -08001182 fput_light(file, fput_needed);
Trond Myklebust6d09bb62006-06-04 02:51:37 -07001183 goto out_fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184}
1185
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001186int fastcall path_lookup(const char *name, unsigned int flags,
1187 struct nameidata *nd)
1188{
1189 return do_path_lookup(AT_FDCWD, name, flags, nd);
1190}
1191
Josef 'Jeff' Sipek16f18202007-07-19 01:48:18 -07001192/**
1193 * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair
1194 * @dentry: pointer to dentry of the base directory
1195 * @mnt: pointer to vfs mount of the base directory
1196 * @name: pointer to file name
1197 * @flags: lookup flags
1198 * @nd: pointer to nameidata
1199 */
1200int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt,
1201 const char *name, unsigned int flags,
1202 struct nameidata *nd)
1203{
1204 int retval;
1205
1206 /* same as do_path_lookup */
1207 nd->last_type = LAST_ROOT;
1208 nd->flags = flags;
1209 nd->depth = 0;
1210
1211 nd->mnt = mntget(mnt);
1212 nd->dentry = dget(dentry);
1213
1214 retval = path_walk(name, nd);
1215 if (unlikely(!retval && !audit_dummy_context() && nd->dentry &&
1216 nd->dentry->d_inode))
Al Viro5a190ae2007-06-07 12:19:32 -04001217 audit_inode(name, nd->dentry);
Josef 'Jeff' Sipek16f18202007-07-19 01:48:18 -07001218
1219 return retval;
1220
1221}
1222
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001223static int __path_lookup_intent_open(int dfd, const char *name,
1224 unsigned int lookup_flags, struct nameidata *nd,
1225 int open_flags, int create_mode)
Trond Myklebust834f2a42005-10-18 14:20:16 -07001226{
1227 struct file *filp = get_empty_filp();
1228 int err;
1229
1230 if (filp == NULL)
1231 return -ENFILE;
1232 nd->intent.open.file = filp;
1233 nd->intent.open.flags = open_flags;
1234 nd->intent.open.create_mode = create_mode;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001235 err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd);
Trond Myklebust834f2a42005-10-18 14:20:16 -07001236 if (IS_ERR(nd->intent.open.file)) {
1237 if (err == 0) {
1238 err = PTR_ERR(nd->intent.open.file);
1239 path_release(nd);
1240 }
1241 } else if (err != 0)
1242 release_open_intent(nd);
1243 return err;
1244}
1245
1246/**
1247 * path_lookup_open - lookup a file path with open intent
Martin Waitz7045f372006-02-01 03:06:57 -08001248 * @dfd: the directory to use as base, or AT_FDCWD
Trond Myklebust834f2a42005-10-18 14:20:16 -07001249 * @name: pointer to file name
1250 * @lookup_flags: lookup intent flags
1251 * @nd: pointer to nameidata
1252 * @open_flags: open intent flags
1253 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001254int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags,
Trond Myklebust834f2a42005-10-18 14:20:16 -07001255 struct nameidata *nd, int open_flags)
1256{
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001257 return __path_lookup_intent_open(dfd, name, lookup_flags, nd,
Trond Myklebust834f2a42005-10-18 14:20:16 -07001258 open_flags, 0);
1259}
1260
1261/**
1262 * path_lookup_create - lookup a file path with open + create intent
Martin Waitz7045f372006-02-01 03:06:57 -08001263 * @dfd: the directory to use as base, or AT_FDCWD
Trond Myklebust834f2a42005-10-18 14:20:16 -07001264 * @name: pointer to file name
1265 * @lookup_flags: lookup intent flags
1266 * @nd: pointer to nameidata
1267 * @open_flags: open intent flags
1268 * @create_mode: create intent flags
1269 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001270static int path_lookup_create(int dfd, const char *name,
1271 unsigned int lookup_flags, struct nameidata *nd,
1272 int open_flags, int create_mode)
Trond Myklebust834f2a42005-10-18 14:20:16 -07001273{
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001274 return __path_lookup_intent_open(dfd, name, lookup_flags|LOOKUP_CREATE,
1275 nd, open_flags, create_mode);
Trond Myklebust834f2a42005-10-18 14:20:16 -07001276}
1277
1278int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags,
1279 struct nameidata *nd, int open_flags)
1280{
1281 char *tmp = getname(name);
1282 int err = PTR_ERR(tmp);
1283
1284 if (!IS_ERR(tmp)) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001285 err = __path_lookup_intent_open(AT_FDCWD, tmp, lookup_flags, nd, open_flags, 0);
Trond Myklebust834f2a42005-10-18 14:20:16 -07001286 putname(tmp);
1287 }
1288 return err;
1289}
1290
Christoph Hellwigeead1912007-10-16 23:25:38 -07001291static struct dentry *__lookup_hash(struct qstr *name,
1292 struct dentry *base, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293{
James Morris057f6c02007-04-26 00:12:05 -07001294 struct dentry *dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 struct inode *inode;
1296 int err;
1297
1298 inode = base->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
1300 /*
1301 * See if the low-level filesystem might want
1302 * to use its own hash..
1303 */
1304 if (base->d_op && base->d_op->d_hash) {
1305 err = base->d_op->d_hash(base, name);
1306 dentry = ERR_PTR(err);
1307 if (err < 0)
1308 goto out;
1309 }
1310
1311 dentry = cached_lookup(base, name, nd);
1312 if (!dentry) {
1313 struct dentry *new = d_alloc(base, name);
1314 dentry = ERR_PTR(-ENOMEM);
1315 if (!new)
1316 goto out;
1317 dentry = inode->i_op->lookup(inode, new, nd);
1318 if (!dentry)
1319 dentry = new;
1320 else
1321 dput(new);
1322 }
1323out:
1324 return dentry;
1325}
1326
James Morris057f6c02007-04-26 00:12:05 -07001327/*
1328 * Restricted form of lookup. Doesn't follow links, single-component only,
1329 * needs parent already locked. Doesn't follow mounts.
1330 * SMP-safe.
1331 */
Adrian Bunka244e162006-03-31 02:32:11 -08001332static struct dentry *lookup_hash(struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333{
Christoph Hellwigeead1912007-10-16 23:25:38 -07001334 int err;
1335
1336 err = permission(nd->dentry->d_inode, MAY_EXEC, nd);
1337 if (err)
1338 return ERR_PTR(err);
Christoph Hellwig49705b72005-11-08 21:35:06 -08001339 return __lookup_hash(&nd->last, nd->dentry, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340}
1341
Christoph Hellwigeead1912007-10-16 23:25:38 -07001342static int __lookup_one_len(const char *name, struct qstr *this,
1343 struct dentry *base, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344{
1345 unsigned long hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 unsigned int c;
1347
James Morris057f6c02007-04-26 00:12:05 -07001348 this->name = name;
1349 this->len = len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 if (!len)
James Morris057f6c02007-04-26 00:12:05 -07001351 return -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
1353 hash = init_name_hash();
1354 while (len--) {
1355 c = *(const unsigned char *)name++;
1356 if (c == '/' || c == '\0')
James Morris057f6c02007-04-26 00:12:05 -07001357 return -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 hash = partial_name_hash(c, hash);
1359 }
James Morris057f6c02007-04-26 00:12:05 -07001360 this->hash = end_name_hash(hash);
1361 return 0;
1362}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
Christoph Hellwigeead1912007-10-16 23:25:38 -07001364/**
1365 * lookup_one_len: filesystem helper to lookup single pathname component
1366 * @name: pathname component to lookup
1367 * @base: base directory to lookup from
1368 * @len: maximum length @len should be interpreted to
1369 *
1370 * Note that this routine is purely a helper for filesystem useage and should
1371 * not be called by generic code. Also note that by using this function to
1372 * nameidata argument is passed to the filesystem methods and a filesystem
1373 * using this helper needs to be prepared for that.
1374 */
James Morris057f6c02007-04-26 00:12:05 -07001375struct dentry *lookup_one_len(const char *name, struct dentry *base, int len)
1376{
1377 int err;
1378 struct qstr this;
1379
1380 err = __lookup_one_len(name, &this, base, len);
1381 if (err)
1382 return ERR_PTR(err);
Christoph Hellwigeead1912007-10-16 23:25:38 -07001383
1384 err = permission(base->d_inode, MAY_EXEC, NULL);
1385 if (err)
1386 return ERR_PTR(err);
Christoph Hellwig49705b72005-11-08 21:35:06 -08001387 return __lookup_hash(&this, base, NULL);
James Morris057f6c02007-04-26 00:12:05 -07001388}
1389
Christoph Hellwigeead1912007-10-16 23:25:38 -07001390/**
1391 * lookup_one_noperm - bad hack for sysfs
1392 * @name: pathname component to lookup
1393 * @base: base directory to lookup from
1394 *
1395 * This is a variant of lookup_one_len that doesn't perform any permission
1396 * checks. It's a horrible hack to work around the braindead sysfs
1397 * architecture and should not be used anywhere else.
1398 *
1399 * DON'T USE THIS FUNCTION EVER, thanks.
1400 */
1401struct dentry *lookup_one_noperm(const char *name, struct dentry *base)
James Morris057f6c02007-04-26 00:12:05 -07001402{
1403 int err;
1404 struct qstr this;
1405
Christoph Hellwigeead1912007-10-16 23:25:38 -07001406 err = __lookup_one_len(name, &this, base, strlen(name));
James Morris057f6c02007-04-26 00:12:05 -07001407 if (err)
1408 return ERR_PTR(err);
Christoph Hellwigeead1912007-10-16 23:25:38 -07001409 return __lookup_hash(&this, base, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410}
1411
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001412int fastcall __user_walk_fd(int dfd, const char __user *name, unsigned flags,
1413 struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414{
1415 char *tmp = getname(name);
1416 int err = PTR_ERR(tmp);
1417
1418 if (!IS_ERR(tmp)) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001419 err = do_path_lookup(dfd, tmp, flags, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 putname(tmp);
1421 }
1422 return err;
1423}
1424
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001425int fastcall __user_walk(const char __user *name, unsigned flags, struct nameidata *nd)
1426{
1427 return __user_walk_fd(AT_FDCWD, name, flags, nd);
1428}
1429
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430/*
1431 * It's inline, so penalty for filesystems that don't use sticky bit is
1432 * minimal.
1433 */
1434static inline int check_sticky(struct inode *dir, struct inode *inode)
1435{
1436 if (!(dir->i_mode & S_ISVTX))
1437 return 0;
1438 if (inode->i_uid == current->fsuid)
1439 return 0;
1440 if (dir->i_uid == current->fsuid)
1441 return 0;
1442 return !capable(CAP_FOWNER);
1443}
1444
1445/*
1446 * Check whether we can remove a link victim from directory dir, check
1447 * whether the type of victim is right.
1448 * 1. We can't do it if dir is read-only (done in permission())
1449 * 2. We should have write and exec permissions on dir
1450 * 3. We can't remove anything from append-only dir
1451 * 4. We can't do anything with immutable dir (done in permission())
1452 * 5. If the sticky bit on dir is set we should either
1453 * a. be owner of dir, or
1454 * b. be owner of victim, or
1455 * c. have CAP_FOWNER capability
1456 * 6. If the victim is append-only or immutable we can't do antyhing with
1457 * links pointing to it.
1458 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
1459 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
1460 * 9. We can't remove a root or mountpoint.
1461 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
1462 * nfs_async_unlink().
1463 */
Arjan van de Ven858119e2006-01-14 13:20:43 -08001464static int may_delete(struct inode *dir,struct dentry *victim,int isdir)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465{
1466 int error;
1467
1468 if (!victim->d_inode)
1469 return -ENOENT;
1470
1471 BUG_ON(victim->d_parent->d_inode != dir);
Al Viro5a190ae2007-06-07 12:19:32 -04001472 audit_inode_child(victim->d_name.name, victim, dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473
1474 error = permission(dir,MAY_WRITE | MAY_EXEC, NULL);
1475 if (error)
1476 return error;
1477 if (IS_APPEND(dir))
1478 return -EPERM;
1479 if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)||
1480 IS_IMMUTABLE(victim->d_inode))
1481 return -EPERM;
1482 if (isdir) {
1483 if (!S_ISDIR(victim->d_inode->i_mode))
1484 return -ENOTDIR;
1485 if (IS_ROOT(victim))
1486 return -EBUSY;
1487 } else if (S_ISDIR(victim->d_inode->i_mode))
1488 return -EISDIR;
1489 if (IS_DEADDIR(dir))
1490 return -ENOENT;
1491 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
1492 return -EBUSY;
1493 return 0;
1494}
1495
1496/* Check whether we can create an object with dentry child in directory
1497 * dir.
1498 * 1. We can't do it if child already exists (open has special treatment for
1499 * this case, but since we are inlined it's OK)
1500 * 2. We can't do it if dir is read-only (done in permission())
1501 * 3. We should have write and exec permissions on dir
1502 * 4. We can't do it if dir is immutable (done in permission())
1503 */
1504static inline int may_create(struct inode *dir, struct dentry *child,
1505 struct nameidata *nd)
1506{
1507 if (child->d_inode)
1508 return -EEXIST;
1509 if (IS_DEADDIR(dir))
1510 return -ENOENT;
1511 return permission(dir,MAY_WRITE | MAY_EXEC, nd);
1512}
1513
1514/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 * O_DIRECTORY translates into forcing a directory lookup.
1516 */
1517static inline int lookup_flags(unsigned int f)
1518{
1519 unsigned long retval = LOOKUP_FOLLOW;
1520
1521 if (f & O_NOFOLLOW)
1522 retval &= ~LOOKUP_FOLLOW;
1523
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 if (f & O_DIRECTORY)
1525 retval |= LOOKUP_DIRECTORY;
1526
1527 return retval;
1528}
1529
1530/*
1531 * p1 and p2 should be directories on the same fs.
1532 */
1533struct dentry *lock_rename(struct dentry *p1, struct dentry *p2)
1534{
1535 struct dentry *p;
1536
1537 if (p1 == p2) {
Ingo Molnarf2eace22006-07-03 00:25:05 -07001538 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 return NULL;
1540 }
1541
Arjan van de Vena11f3a02006-03-23 03:00:33 -08001542 mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543
1544 for (p = p1; p->d_parent != p; p = p->d_parent) {
1545 if (p->d_parent == p2) {
Ingo Molnarf2eace22006-07-03 00:25:05 -07001546 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT);
1547 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 return p;
1549 }
1550 }
1551
1552 for (p = p2; p->d_parent != p; p = p->d_parent) {
1553 if (p->d_parent == p1) {
Ingo Molnarf2eace22006-07-03 00:25:05 -07001554 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1555 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 return p;
1557 }
1558 }
1559
Ingo Molnarf2eace22006-07-03 00:25:05 -07001560 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1561 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 return NULL;
1563}
1564
1565void unlock_rename(struct dentry *p1, struct dentry *p2)
1566{
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001567 mutex_unlock(&p1->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 if (p1 != p2) {
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001569 mutex_unlock(&p2->d_inode->i_mutex);
Arjan van de Vena11f3a02006-03-23 03:00:33 -08001570 mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 }
1572}
1573
1574int vfs_create(struct inode *dir, struct dentry *dentry, int mode,
1575 struct nameidata *nd)
1576{
1577 int error = may_create(dir, dentry, nd);
1578
1579 if (error)
1580 return error;
1581
1582 if (!dir->i_op || !dir->i_op->create)
1583 return -EACCES; /* shouldn't it be ENOSYS? */
1584 mode &= S_IALLUGO;
1585 mode |= S_IFREG;
1586 error = security_inode_create(dir, dentry, mode);
1587 if (error)
1588 return error;
1589 DQUOT_INIT(dir);
1590 error = dir->i_op->create(dir, dentry, mode, nd);
Stephen Smalleya74574a2005-09-09 13:01:44 -07001591 if (!error)
Amy Griffisf38aa942005-11-03 15:57:06 +00001592 fsnotify_create(dir, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 return error;
1594}
1595
1596int may_open(struct nameidata *nd, int acc_mode, int flag)
1597{
1598 struct dentry *dentry = nd->dentry;
1599 struct inode *inode = dentry->d_inode;
1600 int error;
1601
1602 if (!inode)
1603 return -ENOENT;
1604
1605 if (S_ISLNK(inode->i_mode))
1606 return -ELOOP;
1607
Linus Torvalds974a9f02008-01-12 14:06:34 -08001608 if (S_ISDIR(inode->i_mode) && (acc_mode & MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 return -EISDIR;
1610
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 /*
1612 * FIFO's, sockets and device files are special: they don't
1613 * actually live on the filesystem itself, and as such you
1614 * can write to them even if the filesystem is read-only.
1615 */
1616 if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
1617 flag &= ~O_TRUNC;
1618 } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) {
1619 if (nd->mnt->mnt_flags & MNT_NODEV)
1620 return -EACCES;
1621
1622 flag &= ~O_TRUNC;
Linus Torvalds974a9f02008-01-12 14:06:34 -08001623 } else if (IS_RDONLY(inode) && (acc_mode & MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 return -EROFS;
Dave Hansenb41572e2007-10-16 23:31:14 -07001625
1626 error = vfs_permission(nd, acc_mode);
1627 if (error)
1628 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 /*
1630 * An append-only file must be opened in append mode for writing.
1631 */
1632 if (IS_APPEND(inode)) {
1633 if ((flag & FMODE_WRITE) && !(flag & O_APPEND))
1634 return -EPERM;
1635 if (flag & O_TRUNC)
1636 return -EPERM;
1637 }
1638
1639 /* O_NOATIME can only be set by the owner or superuser */
1640 if (flag & O_NOATIME)
Satyam Sharma3bd858a2007-07-17 15:00:08 +05301641 if (!is_owner_or_cap(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 return -EPERM;
1643
1644 /*
1645 * Ensure there are no outstanding leases on the file.
1646 */
1647 error = break_lease(inode, flag);
1648 if (error)
1649 return error;
1650
1651 if (flag & O_TRUNC) {
1652 error = get_write_access(inode);
1653 if (error)
1654 return error;
1655
1656 /*
1657 * Refuse to truncate files with mandatory locks held on them.
1658 */
1659 error = locks_verify_locked(inode);
1660 if (!error) {
1661 DQUOT_INIT(inode);
Miklos Szeredid139d7f2007-10-18 03:07:00 -07001662
1663 error = do_truncate(dentry, 0,
1664 ATTR_MTIME|ATTR_CTIME|ATTR_OPEN,
1665 NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 }
1667 put_write_access(inode);
1668 if (error)
1669 return error;
1670 } else
1671 if (flag & FMODE_WRITE)
1672 DQUOT_INIT(inode);
1673
1674 return 0;
1675}
1676
Dave Hansenaab520e2006-09-30 23:29:02 -07001677static int open_namei_create(struct nameidata *nd, struct path *path,
1678 int flag, int mode)
1679{
1680 int error;
1681 struct dentry *dir = nd->dentry;
1682
1683 if (!IS_POSIXACL(dir->d_inode))
1684 mode &= ~current->fs->umask;
1685 error = vfs_create(dir->d_inode, path->dentry, mode, nd);
1686 mutex_unlock(&dir->d_inode->i_mutex);
1687 dput(nd->dentry);
1688 nd->dentry = path->dentry;
1689 if (error)
1690 return error;
1691 /* Don't check for write permission, don't truncate */
1692 return may_open(nd, 0, flag & ~O_TRUNC);
1693}
1694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695/*
1696 * open_namei()
1697 *
1698 * namei for open - this is in fact almost the whole open-routine.
1699 *
1700 * Note that the low bits of "flag" aren't the same as in the open
1701 * system call - they are 00 - no permissions needed
1702 * 01 - read permission needed
1703 * 10 - write permission needed
1704 * 11 - read/write permissions needed
1705 * which is a lot more logical, and also allows the "no perm" needed
1706 * for symlinks (where the permissions are checked later).
1707 * SMP-safe
1708 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001709int open_namei(int dfd, const char *pathname, int flag,
1710 int mode, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711{
Trond Myklebust834f2a42005-10-18 14:20:16 -07001712 int acc_mode, error;
Al Viro4e7506e2005-06-06 13:36:00 -07001713 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 struct dentry *dir;
1715 int count = 0;
1716
1717 acc_mode = ACC_MODE(flag);
1718
Trond Myklebust834f2a42005-10-18 14:20:16 -07001719 /* O_TRUNC implies we need access checks for write permissions */
1720 if (flag & O_TRUNC)
1721 acc_mode |= MAY_WRITE;
1722
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 /* Allow the LSM permission hook to distinguish append
1724 access from general write access. */
1725 if (flag & O_APPEND)
1726 acc_mode |= MAY_APPEND;
1727
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 /*
1729 * The simplest case - just a plain lookup.
1730 */
1731 if (!(flag & O_CREAT)) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001732 error = path_lookup_open(dfd, pathname, lookup_flags(flag),
1733 nd, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 if (error)
1735 return error;
1736 goto ok;
1737 }
1738
1739 /*
1740 * Create - we need to know the parent.
1741 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001742 error = path_lookup_create(dfd,pathname,LOOKUP_PARENT,nd,flag,mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (error)
1744 return error;
1745
1746 /*
1747 * We have the parent and last component. First of all, check
1748 * that we are not asked to creat(2) an obvious directory - that
1749 * will not do.
1750 */
1751 error = -EISDIR;
1752 if (nd->last_type != LAST_NORM || nd->last.name[nd->last.len])
1753 goto exit;
1754
1755 dir = nd->dentry;
1756 nd->flags &= ~LOOKUP_PARENT;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001757 mutex_lock(&dir->d_inode->i_mutex);
Christoph Hellwig49705b72005-11-08 21:35:06 -08001758 path.dentry = lookup_hash(nd);
Al Virod73ffe12005-06-06 13:36:01 -07001759 path.mnt = nd->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760
1761do_last:
Al Viro4e7506e2005-06-06 13:36:00 -07001762 error = PTR_ERR(path.dentry);
1763 if (IS_ERR(path.dentry)) {
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001764 mutex_unlock(&dir->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 goto exit;
1766 }
1767
Oleg Drokin4af4c522006-03-25 03:06:54 -08001768 if (IS_ERR(nd->intent.open.file)) {
1769 mutex_unlock(&dir->d_inode->i_mutex);
1770 error = PTR_ERR(nd->intent.open.file);
1771 goto exit_dput;
1772 }
1773
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 /* Negative dentry, just create the file */
Al Viro4e7506e2005-06-06 13:36:00 -07001775 if (!path.dentry->d_inode) {
Dave Hansenaab520e2006-09-30 23:29:02 -07001776 error = open_namei_create(nd, &path, flag, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 if (error)
1778 goto exit;
Dave Hansenaab520e2006-09-30 23:29:02 -07001779 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 }
1781
1782 /*
1783 * It already exists.
1784 */
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001785 mutex_unlock(&dir->d_inode->i_mutex);
Al Viro5a190ae2007-06-07 12:19:32 -04001786 audit_inode(pathname, path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787
1788 error = -EEXIST;
1789 if (flag & O_EXCL)
1790 goto exit_dput;
1791
Al Viroe13b2102005-06-06 13:36:06 -07001792 if (__follow_mount(&path)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 error = -ELOOP;
Al Viroba7a4c12005-06-06 13:36:08 -07001794 if (flag & O_NOFOLLOW)
1795 goto exit_dput;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 }
Amy Griffis3e2efce2006-07-13 13:16:02 -04001797
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 error = -ENOENT;
Al Viro4e7506e2005-06-06 13:36:00 -07001799 if (!path.dentry->d_inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 goto exit_dput;
Al Viro4e7506e2005-06-06 13:36:00 -07001801 if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 goto do_link;
1803
Miklos Szeredi09dd17d2005-09-06 15:18:21 -07001804 path_to_nameidata(&path, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 error = -EISDIR;
Al Viro4e7506e2005-06-06 13:36:00 -07001806 if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 goto exit;
1808ok:
1809 error = may_open(nd, acc_mode, flag);
1810 if (error)
1811 goto exit;
1812 return 0;
1813
1814exit_dput:
Miklos Szeredi09dd17d2005-09-06 15:18:21 -07001815 dput_path(&path, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816exit:
Trond Myklebust834f2a42005-10-18 14:20:16 -07001817 if (!IS_ERR(nd->intent.open.file))
1818 release_open_intent(nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 path_release(nd);
1820 return error;
1821
1822do_link:
1823 error = -ELOOP;
1824 if (flag & O_NOFOLLOW)
1825 goto exit_dput;
1826 /*
1827 * This is subtle. Instead of calling do_follow_link() we do the
1828 * thing by hands. The reason is that this way we have zero link_count
1829 * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT.
1830 * After that we have the parent and last component, i.e.
1831 * we are in the same situation as after the first path_walk().
1832 * Well, almost - if the last component is normal we get its copy
1833 * stored in nd->last.name and we will have to putname() it when we
1834 * are done. Procfs-like symlinks just set LAST_BIND.
1835 */
1836 nd->flags |= LOOKUP_PARENT;
Al Viro4e7506e2005-06-06 13:36:00 -07001837 error = security_inode_follow_link(path.dentry, nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 if (error)
1839 goto exit_dput;
Al Virocd4e91d2005-06-06 13:36:03 -07001840 error = __do_follow_link(&path, nd);
Kirill Korotaevde459212006-07-14 00:23:49 -07001841 if (error) {
1842 /* Does someone understand code flow here? Or it is only
1843 * me so stupid? Anathema to whoever designed this non-sense
1844 * with "intent.open".
1845 */
1846 release_open_intent(nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 return error;
Kirill Korotaevde459212006-07-14 00:23:49 -07001848 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 nd->flags &= ~LOOKUP_PARENT;
Al Virod671d5e2005-06-06 13:36:04 -07001850 if (nd->last_type == LAST_BIND)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 goto ok;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 error = -EISDIR;
1853 if (nd->last_type != LAST_NORM)
1854 goto exit;
1855 if (nd->last.name[nd->last.len]) {
Linus Torvalds82984112005-10-06 21:54:21 -07001856 __putname(nd->last.name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 goto exit;
1858 }
1859 error = -ELOOP;
1860 if (count++==32) {
Linus Torvalds82984112005-10-06 21:54:21 -07001861 __putname(nd->last.name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 goto exit;
1863 }
1864 dir = nd->dentry;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001865 mutex_lock(&dir->d_inode->i_mutex);
Christoph Hellwig49705b72005-11-08 21:35:06 -08001866 path.dentry = lookup_hash(nd);
Al Virod671d5e2005-06-06 13:36:04 -07001867 path.mnt = nd->mnt;
Linus Torvalds82984112005-10-06 21:54:21 -07001868 __putname(nd->last.name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 goto do_last;
1870}
1871
1872/**
1873 * lookup_create - lookup a dentry, creating it if it doesn't exist
1874 * @nd: nameidata info
1875 * @is_dir: directory flag
1876 *
1877 * Simple function to lookup and return a dentry and create it
1878 * if it doesn't exist. Is SMP-safe.
Christoph Hellwigc663e5d2005-06-23 00:09:49 -07001879 *
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001880 * Returns with nd->dentry->d_inode->i_mutex locked.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 */
1882struct dentry *lookup_create(struct nameidata *nd, int is_dir)
1883{
Christoph Hellwigc663e5d2005-06-23 00:09:49 -07001884 struct dentry *dentry = ERR_PTR(-EEXIST);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885
Ingo Molnarf2eace22006-07-03 00:25:05 -07001886 mutex_lock_nested(&nd->dentry->d_inode->i_mutex, I_MUTEX_PARENT);
Christoph Hellwigc663e5d2005-06-23 00:09:49 -07001887 /*
1888 * Yucky last component or no last component at all?
1889 * (foo/., foo/.., /////)
1890 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 if (nd->last_type != LAST_NORM)
1892 goto fail;
1893 nd->flags &= ~LOOKUP_PARENT;
ASANO Masahiroa6349042006-08-22 20:06:02 -04001894 nd->flags |= LOOKUP_CREATE;
1895 nd->intent.open.flags = O_EXCL;
Christoph Hellwigc663e5d2005-06-23 00:09:49 -07001896
1897 /*
1898 * Do the final lookup.
1899 */
Christoph Hellwig49705b72005-11-08 21:35:06 -08001900 dentry = lookup_hash(nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 if (IS_ERR(dentry))
1902 goto fail;
Christoph Hellwigc663e5d2005-06-23 00:09:49 -07001903
1904 /*
1905 * Special case - lookup gave negative, but... we had foo/bar/
1906 * From the vfs_mknod() POV we just have a negative dentry -
1907 * all is fine. Let's be bastards - you had / on the end, you've
1908 * been asking for (non-existent) directory. -ENOENT for you.
1909 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
1911 goto enoent;
1912 return dentry;
1913enoent:
1914 dput(dentry);
1915 dentry = ERR_PTR(-ENOENT);
1916fail:
1917 return dentry;
1918}
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -07001919EXPORT_SYMBOL_GPL(lookup_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920
1921int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
1922{
1923 int error = may_create(dir, dentry, NULL);
1924
1925 if (error)
1926 return error;
1927
1928 if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD))
1929 return -EPERM;
1930
1931 if (!dir->i_op || !dir->i_op->mknod)
1932 return -EPERM;
1933
1934 error = security_inode_mknod(dir, dentry, mode, dev);
1935 if (error)
1936 return error;
1937
1938 DQUOT_INIT(dir);
1939 error = dir->i_op->mknod(dir, dentry, mode, dev);
Stephen Smalleya74574a2005-09-09 13:01:44 -07001940 if (!error)
Amy Griffisf38aa942005-11-03 15:57:06 +00001941 fsnotify_create(dir, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 return error;
1943}
1944
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001945asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode,
1946 unsigned dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947{
1948 int error = 0;
1949 char * tmp;
1950 struct dentry * dentry;
1951 struct nameidata nd;
1952
1953 if (S_ISDIR(mode))
1954 return -EPERM;
1955 tmp = getname(filename);
1956 if (IS_ERR(tmp))
1957 return PTR_ERR(tmp);
1958
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001959 error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 if (error)
1961 goto out;
1962 dentry = lookup_create(&nd, 0);
1963 error = PTR_ERR(dentry);
1964
1965 if (!IS_POSIXACL(nd.dentry->d_inode))
1966 mode &= ~current->fs->umask;
1967 if (!IS_ERR(dentry)) {
1968 switch (mode & S_IFMT) {
1969 case 0: case S_IFREG:
1970 error = vfs_create(nd.dentry->d_inode,dentry,mode,&nd);
1971 break;
1972 case S_IFCHR: case S_IFBLK:
1973 error = vfs_mknod(nd.dentry->d_inode,dentry,mode,
1974 new_decode_dev(dev));
1975 break;
1976 case S_IFIFO: case S_IFSOCK:
1977 error = vfs_mknod(nd.dentry->d_inode,dentry,mode,0);
1978 break;
1979 case S_IFDIR:
1980 error = -EPERM;
1981 break;
1982 default:
1983 error = -EINVAL;
1984 }
1985 dput(dentry);
1986 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001987 mutex_unlock(&nd.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 path_release(&nd);
1989out:
1990 putname(tmp);
1991
1992 return error;
1993}
1994
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001995asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev)
1996{
1997 return sys_mknodat(AT_FDCWD, filename, mode, dev);
1998}
1999
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2001{
2002 int error = may_create(dir, dentry, NULL);
2003
2004 if (error)
2005 return error;
2006
2007 if (!dir->i_op || !dir->i_op->mkdir)
2008 return -EPERM;
2009
2010 mode &= (S_IRWXUGO|S_ISVTX);
2011 error = security_inode_mkdir(dir, dentry, mode);
2012 if (error)
2013 return error;
2014
2015 DQUOT_INIT(dir);
2016 error = dir->i_op->mkdir(dir, dentry, mode);
Stephen Smalleya74574a2005-09-09 13:01:44 -07002017 if (!error)
Amy Griffisf38aa942005-11-03 15:57:06 +00002018 fsnotify_mkdir(dir, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 return error;
2020}
2021
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002022asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023{
2024 int error = 0;
2025 char * tmp;
Dave Hansen6902d922006-09-30 23:29:01 -07002026 struct dentry *dentry;
2027 struct nameidata nd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028
2029 tmp = getname(pathname);
2030 error = PTR_ERR(tmp);
Dave Hansen6902d922006-09-30 23:29:01 -07002031 if (IS_ERR(tmp))
2032 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Dave Hansen6902d922006-09-30 23:29:01 -07002034 error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd);
2035 if (error)
2036 goto out;
2037 dentry = lookup_create(&nd, 1);
2038 error = PTR_ERR(dentry);
2039 if (IS_ERR(dentry))
2040 goto out_unlock;
2041
2042 if (!IS_POSIXACL(nd.dentry->d_inode))
2043 mode &= ~current->fs->umask;
2044 error = vfs_mkdir(nd.dentry->d_inode, dentry, mode);
2045 dput(dentry);
2046out_unlock:
2047 mutex_unlock(&nd.dentry->d_inode->i_mutex);
2048 path_release(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049out:
Dave Hansen6902d922006-09-30 23:29:01 -07002050 putname(tmp);
2051out_err:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 return error;
2053}
2054
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002055asmlinkage long sys_mkdir(const char __user *pathname, int mode)
2056{
2057 return sys_mkdirat(AT_FDCWD, pathname, mode);
2058}
2059
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060/*
2061 * We try to drop the dentry early: we should have
2062 * a usage count of 2 if we're the only user of this
2063 * dentry, and if that is true (possibly after pruning
2064 * the dcache), then we drop the dentry now.
2065 *
2066 * A low-level filesystem can, if it choses, legally
2067 * do a
2068 *
2069 * if (!d_unhashed(dentry))
2070 * return -EBUSY;
2071 *
2072 * if it cannot handle the case of removing a directory
2073 * that is still in use by something else..
2074 */
2075void dentry_unhash(struct dentry *dentry)
2076{
2077 dget(dentry);
Vasily Averindc168422006-12-06 20:37:07 -08002078 shrink_dcache_parent(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 spin_lock(&dcache_lock);
2080 spin_lock(&dentry->d_lock);
2081 if (atomic_read(&dentry->d_count) == 2)
2082 __d_drop(dentry);
2083 spin_unlock(&dentry->d_lock);
2084 spin_unlock(&dcache_lock);
2085}
2086
2087int vfs_rmdir(struct inode *dir, struct dentry *dentry)
2088{
2089 int error = may_delete(dir, dentry, 1);
2090
2091 if (error)
2092 return error;
2093
2094 if (!dir->i_op || !dir->i_op->rmdir)
2095 return -EPERM;
2096
2097 DQUOT_INIT(dir);
2098
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002099 mutex_lock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 dentry_unhash(dentry);
2101 if (d_mountpoint(dentry))
2102 error = -EBUSY;
2103 else {
2104 error = security_inode_rmdir(dir, dentry);
2105 if (!error) {
2106 error = dir->i_op->rmdir(dir, dentry);
2107 if (!error)
2108 dentry->d_inode->i_flags |= S_DEAD;
2109 }
2110 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002111 mutex_unlock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 if (!error) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 d_delete(dentry);
2114 }
2115 dput(dentry);
2116
2117 return error;
2118}
2119
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002120static long do_rmdir(int dfd, const char __user *pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121{
2122 int error = 0;
2123 char * name;
2124 struct dentry *dentry;
2125 struct nameidata nd;
2126
2127 name = getname(pathname);
2128 if(IS_ERR(name))
2129 return PTR_ERR(name);
2130
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002131 error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132 if (error)
2133 goto exit;
2134
2135 switch(nd.last_type) {
2136 case LAST_DOTDOT:
2137 error = -ENOTEMPTY;
2138 goto exit1;
2139 case LAST_DOT:
2140 error = -EINVAL;
2141 goto exit1;
2142 case LAST_ROOT:
2143 error = -EBUSY;
2144 goto exit1;
2145 }
Ingo Molnarf2eace22006-07-03 00:25:05 -07002146 mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT);
Christoph Hellwig49705b72005-11-08 21:35:06 -08002147 dentry = lookup_hash(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148 error = PTR_ERR(dentry);
Dave Hansen6902d922006-09-30 23:29:01 -07002149 if (IS_ERR(dentry))
2150 goto exit2;
2151 error = vfs_rmdir(nd.dentry->d_inode, dentry);
2152 dput(dentry);
2153exit2:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002154 mutex_unlock(&nd.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155exit1:
2156 path_release(&nd);
2157exit:
2158 putname(name);
2159 return error;
2160}
2161
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002162asmlinkage long sys_rmdir(const char __user *pathname)
2163{
2164 return do_rmdir(AT_FDCWD, pathname);
2165}
2166
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167int vfs_unlink(struct inode *dir, struct dentry *dentry)
2168{
2169 int error = may_delete(dir, dentry, 0);
2170
2171 if (error)
2172 return error;
2173
2174 if (!dir->i_op || !dir->i_op->unlink)
2175 return -EPERM;
2176
2177 DQUOT_INIT(dir);
2178
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002179 mutex_lock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180 if (d_mountpoint(dentry))
2181 error = -EBUSY;
2182 else {
2183 error = security_inode_unlink(dir, dentry);
2184 if (!error)
2185 error = dir->i_op->unlink(dir, dentry);
2186 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002187 mutex_unlock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188
2189 /* We don't d_delete() NFS sillyrenamed files--they still exist. */
2190 if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) {
Jan Karaece95912008-02-06 01:37:13 -08002191 fsnotify_link_count(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 d_delete(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 }
Robert Love0eeca282005-07-12 17:06:03 -04002194
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 return error;
2196}
2197
2198/*
2199 * Make sure that the actual truncation of the file will occur outside its
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002200 * directory's i_mutex. Truncate can take a long time if there is a lot of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201 * writeout happening, and we don't want to prevent access to the directory
2202 * while waiting on the I/O.
2203 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002204static long do_unlinkat(int dfd, const char __user *pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205{
2206 int error = 0;
2207 char * name;
2208 struct dentry *dentry;
2209 struct nameidata nd;
2210 struct inode *inode = NULL;
2211
2212 name = getname(pathname);
2213 if(IS_ERR(name))
2214 return PTR_ERR(name);
2215
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002216 error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 if (error)
2218 goto exit;
2219 error = -EISDIR;
2220 if (nd.last_type != LAST_NORM)
2221 goto exit1;
Ingo Molnarf2eace22006-07-03 00:25:05 -07002222 mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT);
Christoph Hellwig49705b72005-11-08 21:35:06 -08002223 dentry = lookup_hash(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 error = PTR_ERR(dentry);
2225 if (!IS_ERR(dentry)) {
2226 /* Why not before? Because we want correct error value */
2227 if (nd.last.name[nd.last.len])
2228 goto slashes;
2229 inode = dentry->d_inode;
2230 if (inode)
2231 atomic_inc(&inode->i_count);
2232 error = vfs_unlink(nd.dentry->d_inode, dentry);
2233 exit2:
2234 dput(dentry);
2235 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002236 mutex_unlock(&nd.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237 if (inode)
2238 iput(inode); /* truncate the inode here */
2239exit1:
2240 path_release(&nd);
2241exit:
2242 putname(name);
2243 return error;
2244
2245slashes:
2246 error = !dentry->d_inode ? -ENOENT :
2247 S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR;
2248 goto exit2;
2249}
2250
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002251asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag)
2252{
2253 if ((flag & ~AT_REMOVEDIR) != 0)
2254 return -EINVAL;
2255
2256 if (flag & AT_REMOVEDIR)
2257 return do_rmdir(dfd, pathname);
2258
2259 return do_unlinkat(dfd, pathname);
2260}
2261
2262asmlinkage long sys_unlink(const char __user *pathname)
2263{
2264 return do_unlinkat(AT_FDCWD, pathname);
2265}
2266
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname, int mode)
2268{
2269 int error = may_create(dir, dentry, NULL);
2270
2271 if (error)
2272 return error;
2273
2274 if (!dir->i_op || !dir->i_op->symlink)
2275 return -EPERM;
2276
2277 error = security_inode_symlink(dir, dentry, oldname);
2278 if (error)
2279 return error;
2280
2281 DQUOT_INIT(dir);
2282 error = dir->i_op->symlink(dir, dentry, oldname);
Stephen Smalleya74574a2005-09-09 13:01:44 -07002283 if (!error)
Amy Griffisf38aa942005-11-03 15:57:06 +00002284 fsnotify_create(dir, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002285 return error;
2286}
2287
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002288asmlinkage long sys_symlinkat(const char __user *oldname,
2289 int newdfd, const char __user *newname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290{
2291 int error = 0;
2292 char * from;
2293 char * to;
Dave Hansen6902d922006-09-30 23:29:01 -07002294 struct dentry *dentry;
2295 struct nameidata nd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
2297 from = getname(oldname);
2298 if(IS_ERR(from))
2299 return PTR_ERR(from);
2300 to = getname(newname);
2301 error = PTR_ERR(to);
Dave Hansen6902d922006-09-30 23:29:01 -07002302 if (IS_ERR(to))
2303 goto out_putname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304
Dave Hansen6902d922006-09-30 23:29:01 -07002305 error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd);
2306 if (error)
2307 goto out;
2308 dentry = lookup_create(&nd, 0);
2309 error = PTR_ERR(dentry);
2310 if (IS_ERR(dentry))
2311 goto out_unlock;
2312
2313 error = vfs_symlink(nd.dentry->d_inode, dentry, from, S_IALLUGO);
2314 dput(dentry);
2315out_unlock:
2316 mutex_unlock(&nd.dentry->d_inode->i_mutex);
2317 path_release(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318out:
Dave Hansen6902d922006-09-30 23:29:01 -07002319 putname(to);
2320out_putname:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321 putname(from);
2322 return error;
2323}
2324
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002325asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname)
2326{
2327 return sys_symlinkat(oldname, AT_FDCWD, newname);
2328}
2329
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry)
2331{
2332 struct inode *inode = old_dentry->d_inode;
2333 int error;
2334
2335 if (!inode)
2336 return -ENOENT;
2337
2338 error = may_create(dir, new_dentry, NULL);
2339 if (error)
2340 return error;
2341
2342 if (dir->i_sb != inode->i_sb)
2343 return -EXDEV;
2344
2345 /*
2346 * A link to an append-only or immutable file cannot be created.
2347 */
2348 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2349 return -EPERM;
2350 if (!dir->i_op || !dir->i_op->link)
2351 return -EPERM;
2352 if (S_ISDIR(old_dentry->d_inode->i_mode))
2353 return -EPERM;
2354
2355 error = security_inode_link(old_dentry, dir, new_dentry);
2356 if (error)
2357 return error;
2358
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002359 mutex_lock(&old_dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 DQUOT_INIT(dir);
2361 error = dir->i_op->link(old_dentry, dir, new_dentry);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002362 mutex_unlock(&old_dentry->d_inode->i_mutex);
Stephen Smalleye31e14e2005-09-09 13:01:45 -07002363 if (!error)
Jan Karaece95912008-02-06 01:37:13 -08002364 fsnotify_link(dir, old_dentry->d_inode, new_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365 return error;
2366}
2367
2368/*
2369 * Hardlinks are often used in delicate situations. We avoid
2370 * security-related surprises by not following symlinks on the
2371 * newname. --KAB
2372 *
2373 * We don't follow them on the oldname either to be compatible
2374 * with linux 2.0, and to avoid hard-linking to directories
2375 * and other special files. --ADM
2376 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002377asmlinkage long sys_linkat(int olddfd, const char __user *oldname,
Ulrich Drepperc04030e2006-02-24 13:04:21 -08002378 int newdfd, const char __user *newname,
2379 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380{
2381 struct dentry *new_dentry;
2382 struct nameidata nd, old_nd;
2383 int error;
2384 char * to;
2385
Ulrich Drepper45c9b112006-06-25 05:49:11 -07002386 if ((flags & ~AT_SYMLINK_FOLLOW) != 0)
Ulrich Drepperc04030e2006-02-24 13:04:21 -08002387 return -EINVAL;
2388
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 to = getname(newname);
2390 if (IS_ERR(to))
2391 return PTR_ERR(to);
2392
Ulrich Drepper45c9b112006-06-25 05:49:11 -07002393 error = __user_walk_fd(olddfd, oldname,
2394 flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0,
2395 &old_nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396 if (error)
2397 goto exit;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002398 error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399 if (error)
2400 goto out;
2401 error = -EXDEV;
2402 if (old_nd.mnt != nd.mnt)
2403 goto out_release;
2404 new_dentry = lookup_create(&nd, 0);
2405 error = PTR_ERR(new_dentry);
Dave Hansen6902d922006-09-30 23:29:01 -07002406 if (IS_ERR(new_dentry))
2407 goto out_unlock;
2408 error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry);
2409 dput(new_dentry);
2410out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002411 mutex_unlock(&nd.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412out_release:
2413 path_release(&nd);
2414out:
2415 path_release(&old_nd);
2416exit:
2417 putname(to);
2418
2419 return error;
2420}
2421
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002422asmlinkage long sys_link(const char __user *oldname, const char __user *newname)
2423{
Ulrich Drepperc04030e2006-02-24 13:04:21 -08002424 return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002425}
2426
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427/*
2428 * The worst of all namespace operations - renaming directory. "Perverted"
2429 * doesn't even start to describe it. Somebody in UCB had a heck of a trip...
2430 * Problems:
2431 * a) we can get into loop creation. Check is done in is_subdir().
2432 * b) race potential - two innocent renames can create a loop together.
2433 * That's where 4.4 screws up. Current fix: serialization on
Arjan van de Vena11f3a02006-03-23 03:00:33 -08002434 * sb->s_vfs_rename_mutex. We might be more accurate, but that's another
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 * story.
2436 * c) we have to lock _three_ objects - parents and victim (if it exists).
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002437 * And that - after we got ->i_mutex on parents (until then we don't know
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438 * whether the target exists). Solution: try to be smart with locking
2439 * order for inodes. We rely on the fact that tree topology may change
Arjan van de Vena11f3a02006-03-23 03:00:33 -08002440 * only under ->s_vfs_rename_mutex _and_ that parent of the object we
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 * move will be locked. Thus we can rank directories by the tree
2442 * (ancestors first) and rank all non-directories after them.
2443 * That works since everybody except rename does "lock parent, lookup,
Arjan van de Vena11f3a02006-03-23 03:00:33 -08002444 * lock child" and rename is under ->s_vfs_rename_mutex.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445 * HOWEVER, it relies on the assumption that any object with ->lookup()
2446 * has no more than 1 dentry. If "hybrid" objects will ever appear,
2447 * we'd better make sure that there's no link(2) for them.
2448 * d) some filesystems don't support opened-but-unlinked directories,
2449 * either because of layout or because they are not ready to deal with
2450 * all cases correctly. The latter will be fixed (taking this sort of
2451 * stuff into VFS), but the former is not going away. Solution: the same
2452 * trick as in rmdir().
2453 * e) conversion from fhandle to dentry may come in the wrong moment - when
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002454 * we are removing the target. Solution: we will have to grab ->i_mutex
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002456 * ->i_mutex on parents, which works but leads to some truely excessive
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457 * locking].
2458 */
Adrian Bunk75c96f82005-05-05 16:16:09 -07002459static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry,
2460 struct inode *new_dir, struct dentry *new_dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461{
2462 int error = 0;
2463 struct inode *target;
2464
2465 /*
2466 * If we are going to change the parent - check write permissions,
2467 * we'll need to flip '..'.
2468 */
2469 if (new_dir != old_dir) {
2470 error = permission(old_dentry->d_inode, MAY_WRITE, NULL);
2471 if (error)
2472 return error;
2473 }
2474
2475 error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry);
2476 if (error)
2477 return error;
2478
2479 target = new_dentry->d_inode;
2480 if (target) {
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002481 mutex_lock(&target->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 dentry_unhash(new_dentry);
2483 }
2484 if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry))
2485 error = -EBUSY;
2486 else
2487 error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry);
2488 if (target) {
2489 if (!error)
2490 target->i_flags |= S_DEAD;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002491 mutex_unlock(&target->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002492 if (d_unhashed(new_dentry))
2493 d_rehash(new_dentry);
2494 dput(new_dentry);
2495 }
Stephen Smalleye31e14e2005-09-09 13:01:45 -07002496 if (!error)
Mark Fasheh349457c2006-09-08 14:22:21 -07002497 if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE))
2498 d_move(old_dentry,new_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002499 return error;
2500}
2501
Adrian Bunk75c96f82005-05-05 16:16:09 -07002502static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry,
2503 struct inode *new_dir, struct dentry *new_dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002504{
2505 struct inode *target;
2506 int error;
2507
2508 error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry);
2509 if (error)
2510 return error;
2511
2512 dget(new_dentry);
2513 target = new_dentry->d_inode;
2514 if (target)
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002515 mutex_lock(&target->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry))
2517 error = -EBUSY;
2518 else
2519 error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry);
2520 if (!error) {
Mark Fasheh349457c2006-09-08 14:22:21 -07002521 if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522 d_move(old_dentry, new_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523 }
2524 if (target)
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002525 mutex_unlock(&target->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002526 dput(new_dentry);
2527 return error;
2528}
2529
2530int vfs_rename(struct inode *old_dir, struct dentry *old_dentry,
2531 struct inode *new_dir, struct dentry *new_dentry)
2532{
2533 int error;
2534 int is_dir = S_ISDIR(old_dentry->d_inode->i_mode);
Robert Love0eeca282005-07-12 17:06:03 -04002535 const char *old_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536
2537 if (old_dentry->d_inode == new_dentry->d_inode)
2538 return 0;
2539
2540 error = may_delete(old_dir, old_dentry, is_dir);
2541 if (error)
2542 return error;
2543
2544 if (!new_dentry->d_inode)
2545 error = may_create(new_dir, new_dentry, NULL);
2546 else
2547 error = may_delete(new_dir, new_dentry, is_dir);
2548 if (error)
2549 return error;
2550
2551 if (!old_dir->i_op || !old_dir->i_op->rename)
2552 return -EPERM;
2553
2554 DQUOT_INIT(old_dir);
2555 DQUOT_INIT(new_dir);
2556
Robert Love0eeca282005-07-12 17:06:03 -04002557 old_name = fsnotify_oldname_init(old_dentry->d_name.name);
2558
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559 if (is_dir)
2560 error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry);
2561 else
2562 error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry);
2563 if (!error) {
Robert Love0eeca282005-07-12 17:06:03 -04002564 const char *new_name = old_dentry->d_name.name;
John McCutchan89204c42005-08-15 12:13:28 -04002565 fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir,
Al Viro5a190ae2007-06-07 12:19:32 -04002566 new_dentry->d_inode, old_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567 }
Robert Love0eeca282005-07-12 17:06:03 -04002568 fsnotify_oldname_free(old_name);
2569
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570 return error;
2571}
2572
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002573static int do_rename(int olddfd, const char *oldname,
2574 int newdfd, const char *newname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575{
2576 int error = 0;
2577 struct dentry * old_dir, * new_dir;
2578 struct dentry * old_dentry, *new_dentry;
2579 struct dentry * trap;
2580 struct nameidata oldnd, newnd;
2581
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002582 error = do_path_lookup(olddfd, oldname, LOOKUP_PARENT, &oldnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002583 if (error)
2584 goto exit;
2585
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002586 error = do_path_lookup(newdfd, newname, LOOKUP_PARENT, &newnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587 if (error)
2588 goto exit1;
2589
2590 error = -EXDEV;
2591 if (oldnd.mnt != newnd.mnt)
2592 goto exit2;
2593
2594 old_dir = oldnd.dentry;
2595 error = -EBUSY;
2596 if (oldnd.last_type != LAST_NORM)
2597 goto exit2;
2598
2599 new_dir = newnd.dentry;
2600 if (newnd.last_type != LAST_NORM)
2601 goto exit2;
2602
2603 trap = lock_rename(new_dir, old_dir);
2604
Christoph Hellwig49705b72005-11-08 21:35:06 -08002605 old_dentry = lookup_hash(&oldnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002606 error = PTR_ERR(old_dentry);
2607 if (IS_ERR(old_dentry))
2608 goto exit3;
2609 /* source must exist */
2610 error = -ENOENT;
2611 if (!old_dentry->d_inode)
2612 goto exit4;
2613 /* unless the source is a directory trailing slashes give -ENOTDIR */
2614 if (!S_ISDIR(old_dentry->d_inode->i_mode)) {
2615 error = -ENOTDIR;
2616 if (oldnd.last.name[oldnd.last.len])
2617 goto exit4;
2618 if (newnd.last.name[newnd.last.len])
2619 goto exit4;
2620 }
2621 /* source should not be ancestor of target */
2622 error = -EINVAL;
2623 if (old_dentry == trap)
2624 goto exit4;
Christoph Hellwig49705b72005-11-08 21:35:06 -08002625 new_dentry = lookup_hash(&newnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002626 error = PTR_ERR(new_dentry);
2627 if (IS_ERR(new_dentry))
2628 goto exit4;
2629 /* target should not be an ancestor of source */
2630 error = -ENOTEMPTY;
2631 if (new_dentry == trap)
2632 goto exit5;
2633
2634 error = vfs_rename(old_dir->d_inode, old_dentry,
2635 new_dir->d_inode, new_dentry);
2636exit5:
2637 dput(new_dentry);
2638exit4:
2639 dput(old_dentry);
2640exit3:
2641 unlock_rename(new_dir, old_dir);
2642exit2:
2643 path_release(&newnd);
2644exit1:
2645 path_release(&oldnd);
2646exit:
2647 return error;
2648}
2649
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002650asmlinkage long sys_renameat(int olddfd, const char __user *oldname,
2651 int newdfd, const char __user *newname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652{
2653 int error;
2654 char * from;
2655 char * to;
2656
2657 from = getname(oldname);
2658 if(IS_ERR(from))
2659 return PTR_ERR(from);
2660 to = getname(newname);
2661 error = PTR_ERR(to);
2662 if (!IS_ERR(to)) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002663 error = do_rename(olddfd, from, newdfd, to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002664 putname(to);
2665 }
2666 putname(from);
2667 return error;
2668}
2669
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002670asmlinkage long sys_rename(const char __user *oldname, const char __user *newname)
2671{
2672 return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname);
2673}
2674
Linus Torvalds1da177e2005-04-16 15:20:36 -07002675int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link)
2676{
2677 int len;
2678
2679 len = PTR_ERR(link);
2680 if (IS_ERR(link))
2681 goto out;
2682
2683 len = strlen(link);
2684 if (len > (unsigned) buflen)
2685 len = buflen;
2686 if (copy_to_user(buffer, link, len))
2687 len = -EFAULT;
2688out:
2689 return len;
2690}
2691
2692/*
2693 * A helper for ->readlink(). This should be used *ONLY* for symlinks that
2694 * have ->follow_link() touching nd only in nd_set_link(). Using (or not
2695 * using) it for any given inode is up to filesystem.
2696 */
2697int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen)
2698{
2699 struct nameidata nd;
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002700 void *cookie;
2701
Linus Torvalds1da177e2005-04-16 15:20:36 -07002702 nd.depth = 0;
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002703 cookie = dentry->d_inode->i_op->follow_link(dentry, &nd);
2704 if (!IS_ERR(cookie)) {
2705 int res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706 if (dentry->d_inode->i_op->put_link)
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002707 dentry->d_inode->i_op->put_link(dentry, &nd, cookie);
2708 cookie = ERR_PTR(res);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002709 }
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002710 return PTR_ERR(cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002711}
2712
2713int vfs_follow_link(struct nameidata *nd, const char *link)
2714{
2715 return __vfs_follow_link(nd, link);
2716}
2717
2718/* get the link contents into pagecache */
2719static char *page_getlink(struct dentry * dentry, struct page **ppage)
2720{
2721 struct page * page;
2722 struct address_space *mapping = dentry->d_inode->i_mapping;
Pekka Enberg090d2b12006-06-23 02:05:08 -07002723 page = read_mapping_page(mapping, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002724 if (IS_ERR(page))
Nick Piggin6fe69002007-05-06 14:49:04 -07002725 return (char*)page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726 *ppage = page;
2727 return kmap(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728}
2729
2730int page_readlink(struct dentry *dentry, char __user *buffer, int buflen)
2731{
2732 struct page *page = NULL;
2733 char *s = page_getlink(dentry, &page);
2734 int res = vfs_readlink(dentry,buffer,buflen,s);
2735 if (page) {
2736 kunmap(page);
2737 page_cache_release(page);
2738 }
2739 return res;
2740}
2741
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002742void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002743{
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002744 struct page *page = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002745 nd_set_link(nd, page_getlink(dentry, &page));
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002746 return page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747}
2748
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002749void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002750{
Linus Torvaldscc314ee2005-08-19 18:02:56 -07002751 struct page *page = cookie;
2752
2753 if (page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002754 kunmap(page);
2755 page_cache_release(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756 }
2757}
2758
Kirill Korotaev0adb25d2006-03-11 03:27:13 -08002759int __page_symlink(struct inode *inode, const char *symname, int len,
2760 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761{
2762 struct address_space *mapping = inode->i_mapping;
Kirill Korotaev0adb25d2006-03-11 03:27:13 -08002763 struct page *page;
Nick Pigginafddba42007-10-16 01:25:01 -07002764 void *fsdata;
Dmitriy Monakhovbeb497a2007-02-16 01:27:18 -08002765 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766 char *kaddr;
2767
NeilBrown7e53cac2006-03-25 03:07:57 -08002768retry:
Nick Pigginafddba42007-10-16 01:25:01 -07002769 err = pagecache_write_begin(NULL, mapping, 0, len-1,
2770 AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002771 if (err)
Nick Pigginafddba42007-10-16 01:25:01 -07002772 goto fail;
2773
Linus Torvalds1da177e2005-04-16 15:20:36 -07002774 kaddr = kmap_atomic(page, KM_USER0);
2775 memcpy(kaddr, symname, len-1);
2776 kunmap_atomic(kaddr, KM_USER0);
Nick Pigginafddba42007-10-16 01:25:01 -07002777
2778 err = pagecache_write_end(NULL, mapping, 0, len-1, len-1,
2779 page, fsdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780 if (err < 0)
2781 goto fail;
Nick Pigginafddba42007-10-16 01:25:01 -07002782 if (err < len-1)
2783 goto retry;
2784
Linus Torvalds1da177e2005-04-16 15:20:36 -07002785 mark_inode_dirty(inode);
2786 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002787fail:
2788 return err;
2789}
2790
Kirill Korotaev0adb25d2006-03-11 03:27:13 -08002791int page_symlink(struct inode *inode, const char *symname, int len)
2792{
2793 return __page_symlink(inode, symname, len,
2794 mapping_gfp_mask(inode->i_mapping));
2795}
2796
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -08002797const struct inode_operations page_symlink_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002798 .readlink = generic_readlink,
2799 .follow_link = page_follow_link_light,
2800 .put_link = page_put_link,
2801};
2802
2803EXPORT_SYMBOL(__user_walk);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08002804EXPORT_SYMBOL(__user_walk_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002805EXPORT_SYMBOL(follow_down);
2806EXPORT_SYMBOL(follow_up);
2807EXPORT_SYMBOL(get_write_access); /* binfmt_aout */
2808EXPORT_SYMBOL(getname);
2809EXPORT_SYMBOL(lock_rename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810EXPORT_SYMBOL(lookup_one_len);
2811EXPORT_SYMBOL(page_follow_link_light);
2812EXPORT_SYMBOL(page_put_link);
2813EXPORT_SYMBOL(page_readlink);
Kirill Korotaev0adb25d2006-03-11 03:27:13 -08002814EXPORT_SYMBOL(__page_symlink);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002815EXPORT_SYMBOL(page_symlink);
2816EXPORT_SYMBOL(page_symlink_inode_operations);
2817EXPORT_SYMBOL(path_lookup);
Josef 'Jeff' Sipek16f18202007-07-19 01:48:18 -07002818EXPORT_SYMBOL(vfs_path_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819EXPORT_SYMBOL(path_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002820EXPORT_SYMBOL(permission);
Christoph Hellwige4543ed2005-11-08 21:35:04 -08002821EXPORT_SYMBOL(vfs_permission);
Christoph Hellwig8c744fb2005-11-08 21:35:04 -08002822EXPORT_SYMBOL(file_permission);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002823EXPORT_SYMBOL(unlock_rename);
2824EXPORT_SYMBOL(vfs_create);
2825EXPORT_SYMBOL(vfs_follow_link);
2826EXPORT_SYMBOL(vfs_link);
2827EXPORT_SYMBOL(vfs_mkdir);
2828EXPORT_SYMBOL(vfs_mknod);
2829EXPORT_SYMBOL(generic_permission);
2830EXPORT_SYMBOL(vfs_readlink);
2831EXPORT_SYMBOL(vfs_rename);
2832EXPORT_SYMBOL(vfs_rmdir);
2833EXPORT_SYMBOL(vfs_symlink);
2834EXPORT_SYMBOL(vfs_unlink);
2835EXPORT_SYMBOL(dentry_unhash);
2836EXPORT_SYMBOL(generic_readlink);