projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
fs: introduce vfs_path_lookup
[deliverable/linux.git] / fs / namei.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/namei.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7/*
8 * Some corrections by tytso.
9 */
10
11/* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname
12 * lookup logic.
13 */
14/* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture.
15 */
16
17#include <linux/init.h>
18#include <linux/module.h>
19#include <linux/slab.h>
20#include <linux/fs.h>
21#include <linux/namei.h>
22#include <linux/quotaops.h>
23#include <linux/pagemap.h>
0eeca283 24#include <linux/fsnotify.h>
1da177e4
LT		 25#include <linux/personality.h>
26#include <linux/security.h>
27#include <linux/syscalls.h>
28#include <linux/mount.h>
29#include <linux/audit.h>
16f7e0fe 30#include <linux/capability.h>
834f2a4a 31#include <linux/file.h>
5590ff0d
UD		 32#include <linux/fcntl.h>
33#include <linux/namei.h>
1da177e4
LT		 34#include <asm/namei.h>
35#include <asm/uaccess.h>
36
37#define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE])
38
39/* [Feb-1997 T. Schoebel-Theuer]
40 * Fundamental changes in the pathname lookup mechanisms (namei)
41 * were necessary because of omirr. The reason is that omirr needs
42 * to know the _real_ pathname, not the user-supplied one, in case
43 * of symlinks (and also when transname replacements occur).
44 *
45 * The new code replaces the old recursive symlink resolution with
46 * an iterative one (in case of non-nested symlink chains). It does
47 * this with calls to <fs>_follow_link().
48 * As a side effect, dir_namei(), _namei() and follow_link() are now
49 * replaced with a single function lookup_dentry() that can handle all
50 * the special cases of the former code.
51 *
52 * With the new dcache, the pathname is stored at each inode, at least as
53 * long as the refcount of the inode is positive. As a side effect, the
54 * size of the dcache depends on the inode cache and thus is dynamic.
55 *
56 * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink
57 * resolution to correspond with current state of the code.
58 *
59 * Note that the symlink resolution is not *completely* iterative.
60 * There is still a significant amount of tail- and mid- recursion in
61 * the algorithm. Also, note that <fs>_readlink() is not used in
62 * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink()
63 * may return different results than <fs>_follow_link(). Many virtual
64 * filesystems (including /proc) exhibit this behavior.
65 */
66
67/* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation:
68 * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL
69 * and the name already exists in form of a symlink, try to create the new
70 * name indicated by the symlink. The old code always complained that the
71 * name already exists, due to not following the symlink even if its target
72 * is nonexistent. The new semantics affects also mknod() and link() when
73 * the name is a symlink pointing to a non-existant name.
74 *
75 * I don't know which semantics is the right one, since I have no access
76 * to standards. But I found by trial that HP-UX 9.0 has the full "new"
77 * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the
78 * "old" one. Personally, I think the new semantics is much more logical.
79 * Note that "ln old new" where "new" is a symlink pointing to a non-existing
80 * file does succeed in both HP-UX and SunOs, but not in Solaris
81 * and in the old Linux semantics.
82 */
83
84/* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink
85 * semantics. See the comments in "open_namei" and "do_link" below.
86 *
87 * [10-Sep-98 Alan Modra] Another symlink change.
88 */
89
90/* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks:
91 * inside the path - always follow.
92 * in the last component in creation/removal/renaming - never follow.
93 * if LOOKUP_FOLLOW passed - follow.
94 * if the pathname has trailing slashes - follow.
95 * otherwise - don't follow.
96 * (applied in that order).
97 *
98 * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT
99 * restored for 2.4. This is the last surviving part of old 4.2BSD bug.
100 * During the 2.4 we need to fix the userland stuff depending on it -
101 * hopefully we will be able to get rid of that wart in 2.5. So far only
102 * XEmacs seems to be relying on it...
103 */
104/*
105 * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland)
a11f3a05 106 * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives
1da177e4
LT		 107 * any extra contention...
108 */
109
110/* In order to reduce some races, while at the same time doing additional
111 * checking and hopefully speeding things up, we copy filenames to the
112 * kernel data space before using them..
113 *
114 * POSIX.1 2.4: an empty pathname is invalid (ENOENT).
115 * PATH_MAX includes the nul terminator --RR.
116 */
858119e1 117static int do_getname(const char __user *filename, char *page)
1da177e4
LT		 118{
119 int retval;
120 unsigned long len = PATH_MAX;
121
122 if (!segment_eq(get_fs(), KERNEL_DS)) {
123 if ((unsigned long) filename >= TASK_SIZE)
124 return -EFAULT;
125 if (TASK_SIZE - (unsigned long) filename < PATH_MAX)
126 len = TASK_SIZE - (unsigned long) filename;
127 }
128
129 retval = strncpy_from_user(page, filename, len);
130 if (retval > 0) {
131 if (retval < len)
132 return 0;
133 return -ENAMETOOLONG;
134 } else if (!retval)
135 retval = -ENOENT;
136 return retval;
137}
138
139char * getname(const char __user * filename)
140{
141 char *tmp, *result;
142
143 result = ERR_PTR(-ENOMEM);
144 tmp = __getname();
145 if (tmp) {
146 int retval = do_getname(filename, tmp);
147
148 result = tmp;
149 if (retval < 0) {
150 __putname(tmp);
151 result = ERR_PTR(retval);
152 }
153 }
154 audit_getname(result);
155 return result;
156}
157
158#ifdef CONFIG_AUDITSYSCALL
159void putname(const char *name)
160{
5ac3a9c2 161 if (unlikely(!audit_dummy_context()))
1da177e4
LT		 162 audit_putname(name);
163 else
164 __putname(name);
165}
166EXPORT_SYMBOL(putname);
167#endif
168
169
170/**
171 * generic_permission - check for access rights on a Posix-like filesystem
172 * @inode: inode to check access rights for
173 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
174 * @check_acl: optional callback to check for Posix ACLs
175 *
176 * Used to check for read/write/execute permissions on a file.
177 * We use "fsuid" for this, letting us set arbitrary permissions
178 * for filesystem access without changing the "normal" uids which
179 * are used for other things..
180 */
181int generic_permission(struct inode *inode, int mask,
182 int (*check_acl)(struct inode *inode, int mask))
183{
184 umode_t mode = inode->i_mode;
185
186 if (current->fsuid == inode->i_uid)
187 mode >>= 6;
188 else {
189 if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) {
190 int error = check_acl(inode, mask);
191 if (error == -EACCES)
192 goto check_capabilities;
193 else if (error != -EAGAIN)
194 return error;
195 }
196
197 if (in_group_p(inode->i_gid))
198 mode >>= 3;
199 }
200
201 /*
202 * If the DACs are ok we don't need any capability check.
203 */
204 if (((mode & mask & (MAY_READ|MAY_WRITE|MAY_EXEC)) == mask))
205 return 0;
206
207 check_capabilities:
208 /*
209 * Read/write DACs are always overridable.
210 * Executable DACs are overridable if at least one exec bit is set.
211 */
212 if (!(mask & MAY_EXEC) ||
213 (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode))
214 if (capable(CAP_DAC_OVERRIDE))
215 return 0;
216
217 /*
218 * Searching includes executable on directories, else just read.
219 */
220 if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE)))
221 if (capable(CAP_DAC_READ_SEARCH))
222 return 0;
223
224 return -EACCES;
225}
226
227int permission(struct inode *inode, int mask, struct nameidata *nd)
228{
a343bb77 229 umode_t mode = inode->i_mode;
1da177e4
LT		 230 int retval, submask;
231
232 if (mask & MAY_WRITE) {
1da177e4
LT		 233
234 /*
235 * Nobody gets write access to a read-only fs.
236 */
237 if (IS_RDONLY(inode) &&
238 (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
239 return -EROFS;
240
241 /*
242 * Nobody gets write access to an immutable file.
243 */
244 if (IS_IMMUTABLE(inode))
245 return -EACCES;
246 }
247
248
a343bb77
TM		 249 /*
250 * MAY_EXEC on regular files requires special handling: We override
317a40ac
SS		 251 * filesystem execute permissions if the mode bits aren't set or
252 * the fs is mounted with the "noexec" flag.
a343bb77 253 */
317a40ac
SS		 254 if ((mask & MAY_EXEC) && S_ISREG(mode) && (!(mode & S_IXUGO) ||
255 (nd && nd->mnt && (nd->mnt->mnt_flags & MNT_NOEXEC))))
a343bb77
TM		 256 return -EACCES;
257
1da177e4
LT		 258 /* Ordinary permission routines do not understand MAY_APPEND. */
259 submask = mask & ~MAY_APPEND;
260 if (inode->i_op && inode->i_op->permission)
261 retval = inode->i_op->permission(inode, submask, nd);
262 else
263 retval = generic_permission(inode, submask, NULL);
264 if (retval)
265 return retval;
266
267 return security_inode_permission(inode, mask, nd);
268}
269
e4543edd
CH		 270/**
271 * vfs_permission - check for access rights to a given path
272 * @nd: lookup result that describes the path
273 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
274 *
275 * Used to check for read/write/execute permissions on a path.
276 * We use "fsuid" for this, letting us set arbitrary permissions
277 * for filesystem access without changing the "normal" uids which
278 * are used for other things.
279 */
280int vfs_permission(struct nameidata *nd, int mask)
281{
282 return permission(nd->dentry->d_inode, mask, nd);
283}
284
8c744fb8
CH		 285/**
286 * file_permission - check for additional access rights to a given file
287 * @file: file to check access rights for
288 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
289 *
290 * Used to check for read/write/execute permissions on an already opened
291 * file.
292 *
293 * Note:
294 * Do not use this function in new code. All access checks should
295 * be done using vfs_permission().
296 */
297int file_permission(struct file *file, int mask)
298{
0f7fc9e4 299 return permission(file->f_path.dentry->d_inode, mask, NULL);
8c744fb8
CH		 300}
301
1da177e4
LT		 302/*
303 * get_write_access() gets write permission for a file.
304 * put_write_access() releases this write permission.
305 * This is used for regular files.
306 * We cannot support write (and maybe mmap read-write shared) accesses and
307 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
308 * can have the following values:
309 * 0: no writers, no VM_DENYWRITE mappings
310 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
311 * > 0: (i_writecount) users are writing to the file.
312 *
313 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
314 * except for the cases where we don't hold i_writecount yet. Then we need to
315 * use {get,deny}_write_access() - these functions check the sign and refuse
316 * to do the change if sign is wrong. Exclusion between them is provided by
317 * the inode->i_lock spinlock.
318 */
319
320int get_write_access(struct inode * inode)
321{
322 spin_lock(&inode->i_lock);
323 if (atomic_read(&inode->i_writecount) < 0) {
324 spin_unlock(&inode->i_lock);
325 return -ETXTBSY;
326 }
327 atomic_inc(&inode->i_writecount);
328 spin_unlock(&inode->i_lock);
329
330 return 0;
331}
332
333int deny_write_access(struct file * file)
334{
0f7fc9e4 335 struct inode *inode = file->f_path.dentry->d_inode;
1da177e4
LT		 336
337 spin_lock(&inode->i_lock);
338 if (atomic_read(&inode->i_writecount) > 0) {
339 spin_unlock(&inode->i_lock);
340 return -ETXTBSY;
341 }
342 atomic_dec(&inode->i_writecount);
343 spin_unlock(&inode->i_lock);
344
345 return 0;
346}
347
348void path_release(struct nameidata *nd)
349{
350 dput(nd->dentry);
351 mntput(nd->mnt);
352}
353
354/*
355 * umount() mustn't call path_release()/mntput() as that would clear
356 * mnt_expiry_mark
357 */
358void path_release_on_umount(struct nameidata *nd)
359{
360 dput(nd->dentry);
751c404b 361 mntput_no_expire(nd->mnt);
1da177e4
LT		 362}
363
834f2a4a
TM		 364/**
365 * release_open_intent - free up open intent resources
366 * @nd: pointer to nameidata
367 */
368void release_open_intent(struct nameidata *nd)
369{
0f7fc9e4 370 if (nd->intent.open.file->f_path.dentry == NULL)
834f2a4a
TM		 371 put_filp(nd->intent.open.file);
372 else
373 fput(nd->intent.open.file);
374}
375
bcdc5e01
IK		 376static inline struct dentry *
377do_revalidate(struct dentry *dentry, struct nameidata *nd)
378{
379 int status = dentry->d_op->d_revalidate(dentry, nd);
380 if (unlikely(status <= 0)) {
381 /*
382 * The dentry failed validation.
383 * If d_revalidate returned 0 attempt to invalidate
384 * the dentry otherwise d_revalidate is asking us
385 * to return a fail status.
386 */
387 if (!status) {
388 if (!d_invalidate(dentry)) {
389 dput(dentry);
390 dentry = NULL;
391 }
392 } else {
393 dput(dentry);
394 dentry = ERR_PTR(status);
395 }
396 }
397 return dentry;
398}
399
1da177e4
LT		 400/*
401 * Internal lookup() using the new generic dcache.
402 * SMP-safe
403 */
404static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd)
405{
406 struct dentry * dentry = __d_lookup(parent, name);
407
408 /* lockess __d_lookup may fail due to concurrent d_move()
409 * in some unrelated directory, so try with d_lookup
410 */
411 if (!dentry)
412 dentry = d_lookup(parent, name);
413
bcdc5e01
IK		 414 if (dentry && dentry->d_op && dentry->d_op->d_revalidate)
415 dentry = do_revalidate(dentry, nd);
416
1da177e4
LT		 417 return dentry;
418}
419
420/*
421 * Short-cut version of permission(), for calling by
422 * path_walk(), when dcache lock is held. Combines parts
423 * of permission() and generic_permission(), and tests ONLY for
424 * MAY_EXEC permission.
425 *
426 * If appropriate, check DAC only. If not appropriate, or
427 * short-cut DAC fails, then call permission() to do more
428 * complete permission check.
429 */
858119e1 430static int exec_permission_lite(struct inode *inode,
1da177e4
LT		 431 struct nameidata *nd)
432{
433 umode_t mode = inode->i_mode;
434
435 if (inode->i_op && inode->i_op->permission)
436 return -EAGAIN;
437
438 if (current->fsuid == inode->i_uid)
439 mode >>= 6;
440 else if (in_group_p(inode->i_gid))
441 mode >>= 3;
442
443 if (mode & MAY_EXEC)
444 goto ok;
445
446 if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE))
447 goto ok;
448
449 if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE))
450 goto ok;
451
452 if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH))
453 goto ok;
454
455 return -EACCES;
456ok:
457 return security_inode_permission(inode, MAY_EXEC, nd);
458}
459
460/*
461 * This is called when everything else fails, and we actually have
462 * to go to the low-level filesystem to find out what we should do..
463 *
464 * We get the directory semaphore, and after getting that we also
465 * make sure that nobody added the entry to the dcache in the meantime..
466 * SMP-safe
467 */
468static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd)
469{
470 struct dentry * result;
471 struct inode *dir = parent->d_inode;
472
1b1dcc1b 473 mutex_lock(&dir->i_mutex);
1da177e4
LT		 474 /*
475 * First re-do the cached lookup just in case it was created
476 * while we waited for the directory semaphore..
477 *
478 * FIXME! This could use version numbering or similar to
479 * avoid unnecessary cache lookups.
480 *
481 * The "dcache_lock" is purely to protect the RCU list walker
482 * from concurrent renames at this point (we mustn't get false
483 * negatives from the RCU list walk here, unlike the optimistic
484 * fast walk).
485 *
486 * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup
487 */
488 result = d_lookup(parent, name);
489 if (!result) {
490 struct dentry * dentry = d_alloc(parent, name);
491 result = ERR_PTR(-ENOMEM);
492 if (dentry) {
493 result = dir->i_op->lookup(dir, dentry, nd);
494 if (result)
495 dput(dentry);
496 else
497 result = dentry;
498 }
1b1dcc1b 499 mutex_unlock(&dir->i_mutex);
1da177e4
LT		 500 return result;
501 }
502
503 /*
504 * Uhhuh! Nasty case: the cache was re-populated while
505 * we waited on the semaphore. Need to revalidate.
506 */
1b1dcc1b 507 mutex_unlock(&dir->i_mutex);
1da177e4 508 if (result->d_op && result->d_op->d_revalidate) {
bcdc5e01
IK		 509 result = do_revalidate(result, nd);
510 if (!result)
1da177e4 511 result = ERR_PTR(-ENOENT);
1da177e4
LT		 512 }
513 return result;
514}
515
516static int __emul_lookup_dentry(const char *, struct nameidata *);
517
518/* SMP-safe */
f1662356 519static __always_inline int
1da177e4
LT		 520walk_init_root(const char *name, struct nameidata *nd)
521{
e518ddb7
AM		 522 struct fs_struct *fs = current->fs;
523
524 read_lock(&fs->lock);
525 if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) {
526 nd->mnt = mntget(fs->altrootmnt);
527 nd->dentry = dget(fs->altroot);
528 read_unlock(&fs->lock);
1da177e4
LT		 529 if (__emul_lookup_dentry(name,nd))
530 return 0;
e518ddb7 531 read_lock(&fs->lock);
1da177e4 532 }
e518ddb7
AM		 533 nd->mnt = mntget(fs->rootmnt);
534 nd->dentry = dget(fs->root);
535 read_unlock(&fs->lock);
1da177e4
LT		 536 return 1;
537}
538
f1662356 539static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link)
1da177e4
LT		 540{
541 int res = 0;
542 char *name;
543 if (IS_ERR(link))
544 goto fail;
545
546 if (*link == '/') {
547 path_release(nd);
548 if (!walk_init_root(link, nd))
549 /* weird __emul_prefix() stuff did it */
550 goto out;
551 }
552 res = link_path_walk(link, nd);
553out:
554 if (nd->depth || res || nd->last_type!=LAST_NORM)
555 return res;
556 /*
557 * If it is an iterative symlinks resolution in open_namei() we
558 * have to copy the last component. And all that crap because of
559 * bloody create() on broken symlinks. Furrfu...
560 */
561 name = __getname();
562 if (unlikely(!name)) {
563 path_release(nd);
564 return -ENOMEM;
565 }
566 strcpy(name, nd->last.name);
567 nd->last.name = name;
568 return 0;
569fail:
570 path_release(nd);
571 return PTR_ERR(link);
572}
573
051d3812
IK		 574static inline void dput_path(struct path *path, struct nameidata *nd)
575{
576 dput(path->dentry);
577 if (path->mnt != nd->mnt)
578 mntput(path->mnt);
579}
580
581static inline void path_to_nameidata(struct path *path, struct nameidata *nd)
582{
583 dput(nd->dentry);
584 if (nd->mnt != path->mnt)
585 mntput(nd->mnt);
586 nd->mnt = path->mnt;
587 nd->dentry = path->dentry;
588}
589
f1662356 590static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd)
1da177e4
LT		 591{
592 int error;
cc314eef 593 void *cookie;
cd4e91d3 594 struct dentry *dentry = path->dentry;
1da177e4 595
d671a1cb 596 touch_atime(path->mnt, dentry);
1da177e4 597 nd_set_link(nd, NULL);
cd4e91d3 598
051d3812
IK		 599 if (path->mnt != nd->mnt) {
600 path_to_nameidata(path, nd);
601 dget(dentry);
602 }
603 mntget(path->mnt);
cc314eef
LT		 604 cookie = dentry->d_inode->i_op->follow_link(dentry, nd);
605 error = PTR_ERR(cookie);
606 if (!IS_ERR(cookie)) {
1da177e4 607 char *s = nd_get_link(nd);
cc314eef 608 error = 0;
1da177e4
LT		 609 if (s)
610 error = __vfs_follow_link(nd, s);
611 if (dentry->d_inode->i_op->put_link)
cc314eef 612 dentry->d_inode->i_op->put_link(dentry, nd, cookie);
1da177e4 613 }
cd4e91d3
AV		 614 dput(dentry);
615 mntput(path->mnt);
1da177e4
LT		 616
617 return error;
618}
619
620/*
621 * This limits recursive symlink follows to 8, while
622 * limiting consecutive symlinks to 40.
623 *
624 * Without that kind of total limit, nasty chains of consecutive
625 * symlinks can cause almost arbitrarily long lookups.
626 */
90ebe565 627static inline int do_follow_link(struct path *path, struct nameidata *nd)
1da177e4
LT		 628{
629 int err = -ELOOP;
630 if (current->link_count >= MAX_NESTED_LINKS)
631 goto loop;
632 if (current->total_link_count >= 40)
633 goto loop;
634 BUG_ON(nd->depth >= MAX_NESTED_LINKS);
635 cond_resched();
90ebe565 636 err = security_inode_follow_link(path->dentry, nd);
1da177e4
LT		 637 if (err)
638 goto loop;
639 current->link_count++;
640 current->total_link_count++;
641 nd->depth++;
cd4e91d3 642 err = __do_follow_link(path, nd);
839d9f93
AV		 643 current->link_count--;
644 nd->depth--;
1da177e4
LT		 645 return err;
646loop:
09dd17d3 647 dput_path(path, nd);
839d9f93 648 path_release(nd);
1da177e4
LT		 649 return err;
650}
651
652int follow_up(struct vfsmount **mnt, struct dentry **dentry)
653{
654 struct vfsmount *parent;
655 struct dentry *mountpoint;
656 spin_lock(&vfsmount_lock);
657 parent=(*mnt)->mnt_parent;
658 if (parent == *mnt) {
659 spin_unlock(&vfsmount_lock);
660 return 0;
661 }
662 mntget(parent);
663 mountpoint=dget((*mnt)->mnt_mountpoint);
664 spin_unlock(&vfsmount_lock);
665 dput(*dentry);
666 *dentry = mountpoint;
667 mntput(*mnt);
668 *mnt = parent;
669 return 1;
670}
671
672/* no need for dcache_lock, as serialization is taken care in
673 * namespace.c
674 */
463ffb2e
AV		 675static int __follow_mount(struct path *path)
676{
677 int res = 0;
678 while (d_mountpoint(path->dentry)) {
679 struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry);
680 if (!mounted)
681 break;
682 dput(path->dentry);
683 if (res)
684 mntput(path->mnt);
685 path->mnt = mounted;
686 path->dentry = dget(mounted->mnt_root);
687 res = 1;
688 }
689 return res;
690}
691
58c465eb 692static void follow_mount(struct vfsmount **mnt, struct dentry **dentry)
1da177e4 693{
1da177e4
LT		 694 while (d_mountpoint(*dentry)) {
695 struct vfsmount *mounted = lookup_mnt(*mnt, *dentry);
696 if (!mounted)
697 break;
58c465eb 698 dput(*dentry);
1da177e4
LT		 699 mntput(*mnt);
700 *mnt = mounted;
1da177e4 701 *dentry = dget(mounted->mnt_root);
1da177e4 702 }
1da177e4
LT		 703}
704
705/* no need for dcache_lock, as serialization is taken care in
706 * namespace.c
707 */
e13b210f 708int follow_down(struct vfsmount **mnt, struct dentry **dentry)
1da177e4
LT		 709{
710 struct vfsmount *mounted;
711
712 mounted = lookup_mnt(*mnt, *dentry);
713 if (mounted) {
e13b210f 714 dput(*dentry);
1da177e4
LT		 715 mntput(*mnt);
716 *mnt = mounted;
1da177e4
LT		 717 *dentry = dget(mounted->mnt_root);
718 return 1;
719 }
720 return 0;
721}
722
f1662356 723static __always_inline void follow_dotdot(struct nameidata *nd)
1da177e4 724{
e518ddb7
AM		 725 struct fs_struct *fs = current->fs;
726
1da177e4
LT		 727 while(1) {
728 struct vfsmount *parent;
58c465eb 729 struct dentry *old = nd->dentry;
1da177e4 730
e518ddb7
AM		 731 read_lock(&fs->lock);
732 if (nd->dentry == fs->root &&
733 nd->mnt == fs->rootmnt) {
734 read_unlock(&fs->lock);
1da177e4
LT		 735 break;
736 }
e518ddb7 737 read_unlock(&fs->lock);
1da177e4 738 spin_lock(&dcache_lock);
58c465eb
AV		 739 if (nd->dentry != nd->mnt->mnt_root) {
740 nd->dentry = dget(nd->dentry->d_parent);
1da177e4
LT		 741 spin_unlock(&dcache_lock);
742 dput(old);
743 break;
744 }
745 spin_unlock(&dcache_lock);
746 spin_lock(&vfsmount_lock);
58c465eb
AV		 747 parent = nd->mnt->mnt_parent;
748 if (parent == nd->mnt) {
1da177e4
LT		 749 spin_unlock(&vfsmount_lock);
750 break;
751 }
752 mntget(parent);
58c465eb 753 nd->dentry = dget(nd->mnt->mnt_mountpoint);
1da177e4
LT		 754 spin_unlock(&vfsmount_lock);
755 dput(old);
58c465eb
AV		 756 mntput(nd->mnt);
757 nd->mnt = parent;
1da177e4 758 }
58c465eb 759 follow_mount(&nd->mnt, &nd->dentry);
1da177e4
LT		 760}
761
1da177e4
LT		 762/*
763 * It's more convoluted than I'd like it to be, but... it's still fairly
764 * small and for now I'd prefer to have fast path as straight as possible.
765 * It _is_ time-critical.
766 */
767static int do_lookup(struct nameidata *nd, struct qstr *name,
768 struct path *path)
769{
770 struct vfsmount *mnt = nd->mnt;
771 struct dentry *dentry = __d_lookup(nd->dentry, name);
772
773 if (!dentry)
774 goto need_lookup;
775 if (dentry->d_op && dentry->d_op->d_revalidate)
776 goto need_revalidate;
777done:
778 path->mnt = mnt;
779 path->dentry = dentry;
634ee701 780 __follow_mount(path);
1da177e4
LT		 781 return 0;
782
783need_lookup:
784 dentry = real_lookup(nd->dentry, name, nd);
785 if (IS_ERR(dentry))
786 goto fail;
787 goto done;
788
789need_revalidate:
bcdc5e01
IK		 790 dentry = do_revalidate(dentry, nd);
791 if (!dentry)
792 goto need_lookup;
793 if (IS_ERR(dentry))
794 goto fail;
795 goto done;
1da177e4
LT		 796
797fail:
798 return PTR_ERR(dentry);
799}
800
801/*
802 * Name resolution.
ea3834d9
PM		 803 * This is the basic name resolution function, turning a pathname into
804 * the final dentry. We expect 'base' to be positive and a directory.
1da177e4 805 *
ea3834d9
PM		 806 * Returns 0 and nd will have valid dentry and mnt on success.
807 * Returns error and drops reference to input namei data on failure.
1da177e4
LT		 808 */
809static fastcall int __link_path_walk(const char * name, struct nameidata *nd)
810{
811 struct path next;
812 struct inode *inode;
813 int err;
814 unsigned int lookup_flags = nd->flags;
815
816 while (*name=='/')
817 name++;
818 if (!*name)
819 goto return_reval;
820
821 inode = nd->dentry->d_inode;
822 if (nd->depth)
f55eab82 823 lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE);
1da177e4
LT		 824
825 /* At this point we know we have a real path component. */
826 for(;;) {
827 unsigned long hash;
828 struct qstr this;
829 unsigned int c;
830
cdce5d6b 831 nd->flags |= LOOKUP_CONTINUE;
1da177e4 832 err = exec_permission_lite(inode, nd);
e4543edd
CH		 833 if (err == -EAGAIN)
834 err = vfs_permission(nd, MAY_EXEC);
1da177e4
LT		 835 if (err)
836 break;
837
838 this.name = name;
839 c = *(const unsigned char *)name;
840
841 hash = init_name_hash();
842 do {
843 name++;
844 hash = partial_name_hash(c, hash);
845 c = *(const unsigned char *)name;
846 } while (c && (c != '/'));
847 this.len = name - (const char *) this.name;
848 this.hash = end_name_hash(hash);
849
850 /* remove trailing slashes? */
851 if (!c)
852 goto last_component;
853 while (*++name == '/');
854 if (!*name)
855 goto last_with_slashes;
856
857 /*
858 * "." and ".." are special - ".." especially so because it has
859 * to be able to know about the current root directory and
860 * parent relationships.
861 */
862 if (this.name[0] == '.') switch (this.len) {
863 default:
864 break;
865 case 2:
866 if (this.name[1] != '.')
867 break;
58c465eb 868 follow_dotdot(nd);
1da177e4
LT		 869 inode = nd->dentry->d_inode;
870 /* fallthrough */
871 case 1:
872 continue;
873 }
874 /*
875 * See if the low-level filesystem might want
876 * to use its own hash..
877 */
878 if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
879 err = nd->dentry->d_op->d_hash(nd->dentry, &this);
880 if (err < 0)
881 break;
882 }
1da177e4
LT		 883 /* This does the actual lookups.. */
884 err = do_lookup(nd, &this, &next);
885 if (err)
886 break;
1da177e4
LT		 887
888 err = -ENOENT;
889 inode = next.dentry->d_inode;
890 if (!inode)
891 goto out_dput;
892 err = -ENOTDIR;
893 if (!inode->i_op)
894 goto out_dput;
895
896 if (inode->i_op->follow_link) {
90ebe565 897 err = do_follow_link(&next, nd);
1da177e4
LT		 898 if (err)
899 goto return_err;
900 err = -ENOENT;
901 inode = nd->dentry->d_inode;
902 if (!inode)
903 break;
904 err = -ENOTDIR;
905 if (!inode->i_op)
906 break;
09dd17d3
MS		 907 } else
908 path_to_nameidata(&next, nd);
1da177e4
LT		 909 err = -ENOTDIR;
910 if (!inode->i_op->lookup)
911 break;
912 continue;
913 /* here ends the main loop */
914
915last_with_slashes:
916 lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
917last_component:
f55eab82
TM		 918 /* Clear LOOKUP_CONTINUE iff it was previously unset */
919 nd->flags &= lookup_flags | ~LOOKUP_CONTINUE;
1da177e4
LT		 920 if (lookup_flags & LOOKUP_PARENT)
921 goto lookup_parent;
922 if (this.name[0] == '.') switch (this.len) {
923 default:
924 break;
925 case 2:
926 if (this.name[1] != '.')
927 break;
58c465eb 928 follow_dotdot(nd);
1da177e4
LT		 929 inode = nd->dentry->d_inode;
930 /* fallthrough */
931 case 1:
932 goto return_reval;
933 }
934 if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
935 err = nd->dentry->d_op->d_hash(nd->dentry, &this);
936 if (err < 0)
937 break;
938 }
939 err = do_lookup(nd, &this, &next);
940 if (err)
941 break;
1da177e4
LT		 942 inode = next.dentry->d_inode;
943 if ((lookup_flags & LOOKUP_FOLLOW)
944 && inode && inode->i_op && inode->i_op->follow_link) {
90ebe565 945 err = do_follow_link(&next, nd);
1da177e4
LT		 946 if (err)
947 goto return_err;
948 inode = nd->dentry->d_inode;
09dd17d3
MS		 949 } else
950 path_to_nameidata(&next, nd);
1da177e4
LT		 951 err = -ENOENT;
952 if (!inode)
953 break;
954 if (lookup_flags & LOOKUP_DIRECTORY) {
955 err = -ENOTDIR;
956 if (!inode->i_op || !inode->i_op->lookup)
957 break;
958 }
959 goto return_base;
960lookup_parent:
961 nd->last = this;
962 nd->last_type = LAST_NORM;
963 if (this.name[0] != '.')
964 goto return_base;
965 if (this.len == 1)
966 nd->last_type = LAST_DOT;
967 else if (this.len == 2 && this.name[1] == '.')
968 nd->last_type = LAST_DOTDOT;
969 else
970 goto return_base;
971return_reval:
972 /*
973 * We bypassed the ordinary revalidation routines.
974 * We may need to check the cached dentry for staleness.
975 */
976 if (nd->dentry && nd->dentry->d_sb &&
977 (nd->dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) {
978 err = -ESTALE;
979 /* Note: we do not d_invalidate() */
980 if (!nd->dentry->d_op->d_revalidate(nd->dentry, nd))
981 break;
982 }
983return_base:
984 return 0;
985out_dput:
09dd17d3 986 dput_path(&next, nd);
1da177e4
LT		 987 break;
988 }
989 path_release(nd);
990return_err:
991 return err;
992}
993
994/*
995 * Wrapper to retry pathname resolution whenever the underlying
996 * file system returns an ESTALE.
997 *
998 * Retry the whole path once, forcing real lookup requests
999 * instead of relying on the dcache.
1000 */
1001int fastcall link_path_walk(const char *name, struct nameidata *nd)
1002{
1003 struct nameidata save = *nd;
1004 int result;
1005
1006 /* make sure the stuff we saved doesn't go away */
1007 dget(save.dentry);
1008 mntget(save.mnt);
1009
1010 result = __link_path_walk(name, nd);
1011 if (result == -ESTALE) {
1012 *nd = save;
1013 dget(nd->dentry);
1014 mntget(nd->mnt);
1015 nd->flags |= LOOKUP_REVAL;
1016 result = __link_path_walk(name, nd);
1017 }
1018
1019 dput(save.dentry);
1020 mntput(save.mnt);
1021
1022 return result;
1023}
1024
1025int fastcall path_walk(const char * name, struct nameidata *nd)
1026{
1027 current->total_link_count = 0;
1028 return link_path_walk(name, nd);
1029}
1030
ea3834d9
PM		 1031/*
1032 * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if
1033 * everything is done. Returns 0 and drops input nd, if lookup failed;
1034 */
1da177e4
LT		 1035static int __emul_lookup_dentry(const char *name, struct nameidata *nd)
1036{
1037 if (path_walk(name, nd))
1038 return 0; /* something went wrong... */
1039
1040 if (!nd->dentry->d_inode || S_ISDIR(nd->dentry->d_inode->i_mode)) {
1041 struct dentry *old_dentry = nd->dentry;
1042 struct vfsmount *old_mnt = nd->mnt;
1043 struct qstr last = nd->last;
1044 int last_type = nd->last_type;
e518ddb7
AM		 1045 struct fs_struct *fs = current->fs;
1046
1da177e4 1047 /*
e518ddb7
AM		 1048 * NAME was not found in alternate root or it's a directory.
1049 * Try to find it in the normal root:
1da177e4
LT		 1050 */
1051 nd->last_type = LAST_ROOT;
e518ddb7
AM		 1052 read_lock(&fs->lock);
1053 nd->mnt = mntget(fs->rootmnt);
1054 nd->dentry = dget(fs->root);
1055 read_unlock(&fs->lock);
1da177e4
LT		 1056 if (path_walk(name, nd) == 0) {
1057 if (nd->dentry->d_inode) {
1058 dput(old_dentry);
1059 mntput(old_mnt);
1060 return 1;
1061 }
1062 path_release(nd);
1063 }
1064 nd->dentry = old_dentry;
1065 nd->mnt = old_mnt;
1066 nd->last = last;
1067 nd->last_type = last_type;
1068 }
1069 return 1;
1070}
1071
1072void set_fs_altroot(void)
1073{
1074 char *emul = __emul_prefix();
1075 struct nameidata nd;
1076 struct vfsmount *mnt = NULL, *oldmnt;
1077 struct dentry *dentry = NULL, *olddentry;
1078 int err;
e518ddb7 1079 struct fs_struct *fs = current->fs;
1da177e4
LT		 1080
1081 if (!emul)
1082 goto set_it;
1083 err = path_lookup(emul, LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_NOALT, &nd);
1084 if (!err) {
1085 mnt = nd.mnt;
1086 dentry = nd.dentry;
1087 }
1088set_it:
e518ddb7
AM		 1089 write_lock(&fs->lock);
1090 oldmnt = fs->altrootmnt;
1091 olddentry = fs->altroot;
1092 fs->altrootmnt = mnt;
1093 fs->altroot = dentry;
1094 write_unlock(&fs->lock);
1da177e4
LT		 1095 if (olddentry) {
1096 dput(olddentry);
1097 mntput(oldmnt);
1098 }
1099}
1100
ea3834d9 1101/* Returns 0 and nd will be valid on success; Retuns error, otherwise. */
5590ff0d
UD		 1102static int fastcall do_path_lookup(int dfd, const char *name,
1103 unsigned int flags, struct nameidata *nd)
1da177e4 1104{
ea3834d9 1105 int retval = 0;
170aa3d0
UD		 1106 int fput_needed;
1107 struct file *file;
e518ddb7 1108 struct fs_struct *fs = current->fs;
1da177e4
LT		 1109
1110 nd->last_type = LAST_ROOT; /* if there are only slashes... */
1111 nd->flags = flags;
1112 nd->depth = 0;
1113
1da177e4 1114 if (*name=='/') {
e518ddb7
AM		 1115 read_lock(&fs->lock);
1116 if (fs->altroot && !(nd->flags & LOOKUP_NOALT)) {
1117 nd->mnt = mntget(fs->altrootmnt);
1118 nd->dentry = dget(fs->altroot);
1119 read_unlock(&fs->lock);
1da177e4 1120 if (__emul_lookup_dentry(name,nd))
ea3834d9 1121 goto out; /* found in altroot */
e518ddb7 1122 read_lock(&fs->lock);
1da177e4 1123 }
e518ddb7
AM		 1124 nd->mnt = mntget(fs->rootmnt);
1125 nd->dentry = dget(fs->root);
1126 read_unlock(&fs->lock);
5590ff0d 1127 } else if (dfd == AT_FDCWD) {
e518ddb7
AM		 1128 read_lock(&fs->lock);
1129 nd->mnt = mntget(fs->pwdmnt);
1130 nd->dentry = dget(fs->pwd);
1131 read_unlock(&fs->lock);
5590ff0d 1132 } else {
5590ff0d
UD		 1133 struct dentry *dentry;
1134
1135 file = fget_light(dfd, &fput_needed);
170aa3d0
UD		 1136 retval = -EBADF;
1137 if (!file)
6d09bb62 1138 goto out_fail;
5590ff0d 1139
0f7fc9e4 1140 dentry = file->f_path.dentry;
5590ff0d 1141
170aa3d0
UD		 1142 retval = -ENOTDIR;
1143 if (!S_ISDIR(dentry->d_inode->i_mode))
6d09bb62 1144 goto fput_fail;
5590ff0d
UD		 1145
1146 retval = file_permission(file, MAY_EXEC);
170aa3d0 1147 if (retval)
6d09bb62 1148 goto fput_fail;
5590ff0d 1149
0f7fc9e4 1150 nd->mnt = mntget(file->f_path.mnt);
5590ff0d
UD		 1151 nd->dentry = dget(dentry);
1152
1153 fput_light(file, fput_needed);
1da177e4 1154 }
2dfdd266
JJS		 1155
1156 retval = path_walk(name, nd);
ea3834d9 1157out:
62ce39c5 1158 if (unlikely(!retval && !audit_dummy_context() && nd->dentry &&
3bc8414b 1159 nd->dentry->d_inode))
9c937dcc 1160 audit_inode(name, nd->dentry->d_inode);
6d09bb62 1161out_fail:
170aa3d0
UD		 1162 return retval;
1163
6d09bb62 1164fput_fail:
170aa3d0 1165 fput_light(file, fput_needed);
6d09bb62 1166 goto out_fail;
1da177e4
LT		 1167}
1168
5590ff0d
UD		 1169int fastcall path_lookup(const char *name, unsigned int flags,
1170 struct nameidata *nd)
1171{
1172 return do_path_lookup(AT_FDCWD, name, flags, nd);
1173}
1174
16f18200
JJS		 1175/**
1176 * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair
1177 * @dentry: pointer to dentry of the base directory
1178 * @mnt: pointer to vfs mount of the base directory
1179 * @name: pointer to file name
1180 * @flags: lookup flags
1181 * @nd: pointer to nameidata
1182 */
1183int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt,
1184 const char *name, unsigned int flags,
1185 struct nameidata *nd)
1186{
1187 int retval;
1188
1189 /* same as do_path_lookup */
1190 nd->last_type = LAST_ROOT;
1191 nd->flags = flags;
1192 nd->depth = 0;
1193
1194 nd->mnt = mntget(mnt);
1195 nd->dentry = dget(dentry);
1196
1197 retval = path_walk(name, nd);
1198 if (unlikely(!retval && !audit_dummy_context() && nd->dentry &&
1199 nd->dentry->d_inode))
1200 audit_inode(name, nd->dentry->d_inode);
1201
1202 return retval;
1203
1204}
1205
5590ff0d
UD		 1206static int __path_lookup_intent_open(int dfd, const char *name,
1207 unsigned int lookup_flags, struct nameidata *nd,
1208 int open_flags, int create_mode)
834f2a4a
TM		 1209{
1210 struct file *filp = get_empty_filp();
1211 int err;
1212
1213 if (filp == NULL)
1214 return -ENFILE;
1215 nd->intent.open.file = filp;
1216 nd->intent.open.flags = open_flags;
1217 nd->intent.open.create_mode = create_mode;
5590ff0d 1218 err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd);
834f2a4a
TM		 1219 if (IS_ERR(nd->intent.open.file)) {
1220 if (err == 0) {
1221 err = PTR_ERR(nd->intent.open.file);
1222 path_release(nd);
1223 }
1224 } else if (err != 0)
1225 release_open_intent(nd);
1226 return err;
1227}
1228
1229/**
1230 * path_lookup_open - lookup a file path with open intent
7045f37b 1231 * @dfd: the directory to use as base, or AT_FDCWD
834f2a4a
TM		 1232 * @name: pointer to file name
1233 * @lookup_flags: lookup intent flags
1234 * @nd: pointer to nameidata
1235 * @open_flags: open intent flags
1236 */
5590ff0d 1237int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags,
834f2a4a
TM		 1238 struct nameidata *nd, int open_flags)
1239{
5590ff0d 1240 return __path_lookup_intent_open(dfd, name, lookup_flags, nd,
834f2a4a
TM		 1241 open_flags, 0);
1242}
1243
1244/**
1245 * path_lookup_create - lookup a file path with open + create intent
7045f37b 1246 * @dfd: the directory to use as base, or AT_FDCWD
834f2a4a
TM		 1247 * @name: pointer to file name
1248 * @lookup_flags: lookup intent flags
1249 * @nd: pointer to nameidata
1250 * @open_flags: open intent flags
1251 * @create_mode: create intent flags
1252 */
5590ff0d
UD		 1253static int path_lookup_create(int dfd, const char *name,
1254 unsigned int lookup_flags, struct nameidata *nd,
1255 int open_flags, int create_mode)
834f2a4a 1256{
5590ff0d
UD		 1257 return __path_lookup_intent_open(dfd, name, lookup_flags|LOOKUP_CREATE,
1258 nd, open_flags, create_mode);
834f2a4a
TM		 1259}
1260
1261int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags,
1262 struct nameidata *nd, int open_flags)
1263{
1264 char *tmp = getname(name);
1265 int err = PTR_ERR(tmp);
1266
1267 if (!IS_ERR(tmp)) {
5590ff0d 1268 err = __path_lookup_intent_open(AT_FDCWD, tmp, lookup_flags, nd, open_flags, 0);
834f2a4a
TM		 1269 putname(tmp);
1270 }
1271 return err;
1272}
1273
057f6c01 1274static inline struct dentry *__lookup_hash_kern(struct qstr *name, struct dentry *base, struct nameidata *nd)
1da177e4 1275{
057f6c01 1276 struct dentry *dentry;
1da177e4
LT		 1277 struct inode *inode;
1278 int err;
1279
1280 inode = base->d_inode;
1da177e4
LT		 1281
1282 /*
1283 * See if the low-level filesystem might want
1284 * to use its own hash..
1285 */
1286 if (base->d_op && base->d_op->d_hash) {
1287 err = base->d_op->d_hash(base, name);
1288 dentry = ERR_PTR(err);
1289 if (err < 0)
1290 goto out;
1291 }
1292
1293 dentry = cached_lookup(base, name, nd);
1294 if (!dentry) {
1295 struct dentry *new = d_alloc(base, name);
1296 dentry = ERR_PTR(-ENOMEM);
1297 if (!new)
1298 goto out;
1299 dentry = inode->i_op->lookup(inode, new, nd);
1300 if (!dentry)
1301 dentry = new;
1302 else
1303 dput(new);
1304 }
1305out:
1306 return dentry;
1307}
1308
057f6c01
JM		 1309/*
1310 * Restricted form of lookup. Doesn't follow links, single-component only,
1311 * needs parent already locked. Doesn't follow mounts.
1312 * SMP-safe.
1313 */
1314static inline struct dentry * __lookup_hash(struct qstr *name, struct dentry *base, struct nameidata *nd)
1315{
1316 struct dentry *dentry;
1317 struct inode *inode;
1318 int err;
1319
1320 inode = base->d_inode;
1321
1322 err = permission(inode, MAY_EXEC, nd);
1323 dentry = ERR_PTR(err);
1324 if (err)
1325 goto out;
1326
1327 dentry = __lookup_hash_kern(name, base, nd);
1328out:
1329 return dentry;
1330}
1331
a244e169 1332static struct dentry *lookup_hash(struct nameidata *nd)
1da177e4 1333{
49705b77 1334 return __lookup_hash(&nd->last, nd->dentry, nd);
1da177e4
LT		 1335}
1336
1337/* SMP-safe */
057f6c01 1338static inline int __lookup_one_len(const char *name, struct qstr *this, struct dentry *base, int len)
1da177e4
LT		 1339{
1340 unsigned long hash;
1da177e4
LT		 1341 unsigned int c;
1342
057f6c01
JM		 1343 this->name = name;
1344 this->len = len;
1da177e4 1345 if (!len)
057f6c01 1346 return -EACCES;
1da177e4
LT		 1347
1348 hash = init_name_hash();
1349 while (len--) {
1350 c = *(const unsigned char *)name++;
1351 if (c == '/' || c == '\0')
057f6c01 1352 return -EACCES;
1da177e4
LT		 1353 hash = partial_name_hash(c, hash);
1354 }
057f6c01
JM		 1355 this->hash = end_name_hash(hash);
1356 return 0;
1357}
1da177e4 1358
057f6c01
JM		 1359struct dentry *lookup_one_len(const char *name, struct dentry *base, int len)
1360{
1361 int err;
1362 struct qstr this;
1363
1364 err = __lookup_one_len(name, &this, base, len);
1365 if (err)
1366 return ERR_PTR(err);
49705b77 1367 return __lookup_hash(&this, base, NULL);
057f6c01
JM		 1368}
1369
1370struct dentry *lookup_one_len_kern(const char *name, struct dentry *base, int len)
1371{
1372 int err;
1373 struct qstr this;
1374
1375 err = __lookup_one_len(name, &this, base, len);
1376 if (err)
1377 return ERR_PTR(err);
1378 return __lookup_hash_kern(&this, base, NULL);
1da177e4
LT		 1379}
1380
5590ff0d
UD		 1381int fastcall __user_walk_fd(int dfd, const char __user *name, unsigned flags,
1382 struct nameidata *nd)
1da177e4
LT		 1383{
1384 char *tmp = getname(name);
1385 int err = PTR_ERR(tmp);
1386
1387 if (!IS_ERR(tmp)) {
5590ff0d 1388 err = do_path_lookup(dfd, tmp, flags, nd);
1da177e4
LT		 1389 putname(tmp);
1390 }
1391 return err;
1392}
1393
5590ff0d
UD		 1394int fastcall __user_walk(const char __user *name, unsigned flags, struct nameidata *nd)
1395{
1396 return __user_walk_fd(AT_FDCWD, name, flags, nd);
1397}
1398
1da177e4
LT		 1399/*
1400 * It's inline, so penalty for filesystems that don't use sticky bit is
1401 * minimal.
1402 */
1403static inline int check_sticky(struct inode *dir, struct inode *inode)
1404{
1405 if (!(dir->i_mode & S_ISVTX))
1406 return 0;
1407 if (inode->i_uid == current->fsuid)
1408 return 0;
1409 if (dir->i_uid == current->fsuid)
1410 return 0;
1411 return !capable(CAP_FOWNER);
1412}
1413
1414/*
1415 * Check whether we can remove a link victim from directory dir, check
1416 * whether the type of victim is right.
1417 * 1. We can't do it if dir is read-only (done in permission())
1418 * 2. We should have write and exec permissions on dir
1419 * 3. We can't remove anything from append-only dir
1420 * 4. We can't do anything with immutable dir (done in permission())
1421 * 5. If the sticky bit on dir is set we should either
1422 * a. be owner of dir, or
1423 * b. be owner of victim, or
1424 * c. have CAP_FOWNER capability
1425 * 6. If the victim is append-only or immutable we can't do antyhing with
1426 * links pointing to it.
1427 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
1428 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
1429 * 9. We can't remove a root or mountpoint.
1430 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
1431 * nfs_async_unlink().
1432 */
858119e1 1433static int may_delete(struct inode *dir,struct dentry *victim,int isdir)
1da177e4
LT		 1434{
1435 int error;
1436
1437 if (!victim->d_inode)
1438 return -ENOENT;
1439
1440 BUG_ON(victim->d_parent->d_inode != dir);
73d3ec5a 1441 audit_inode_child(victim->d_name.name, victim->d_inode, dir);
1da177e4
LT		 1442
1443 error = permission(dir,MAY_WRITE | MAY_EXEC, NULL);
1444 if (error)
1445 return error;
1446 if (IS_APPEND(dir))
1447 return -EPERM;
1448 if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)||
1449 IS_IMMUTABLE(victim->d_inode))
1450 return -EPERM;
1451 if (isdir) {
1452 if (!S_ISDIR(victim->d_inode->i_mode))
1453 return -ENOTDIR;
1454 if (IS_ROOT(victim))
1455 return -EBUSY;
1456 } else if (S_ISDIR(victim->d_inode->i_mode))
1457 return -EISDIR;
1458 if (IS_DEADDIR(dir))
1459 return -ENOENT;
1460 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
1461 return -EBUSY;
1462 return 0;
1463}
1464
1465/* Check whether we can create an object with dentry child in directory
1466 * dir.
1467 * 1. We can't do it if child already exists (open has special treatment for
1468 * this case, but since we are inlined it's OK)
1469 * 2. We can't do it if dir is read-only (done in permission())
1470 * 3. We should have write and exec permissions on dir
1471 * 4. We can't do it if dir is immutable (done in permission())
1472 */
1473static inline int may_create(struct inode *dir, struct dentry *child,
1474 struct nameidata *nd)
1475{
1476 if (child->d_inode)
1477 return -EEXIST;
1478 if (IS_DEADDIR(dir))
1479 return -ENOENT;
1480 return permission(dir,MAY_WRITE | MAY_EXEC, nd);
1481}
1482
1483/*
1da177e4
LT		 1484 * O_DIRECTORY translates into forcing a directory lookup.
1485 */
1486static inline int lookup_flags(unsigned int f)
1487{
1488 unsigned long retval = LOOKUP_FOLLOW;
1489
1490 if (f & O_NOFOLLOW)
1491 retval &= ~LOOKUP_FOLLOW;
1492
1da177e4
LT		 1493 if (f & O_DIRECTORY)
1494 retval |= LOOKUP_DIRECTORY;
1495
1496 return retval;
1497}
1498
1499/*
1500 * p1 and p2 should be directories on the same fs.
1501 */
1502struct dentry *lock_rename(struct dentry *p1, struct dentry *p2)
1503{
1504 struct dentry *p;
1505
1506 if (p1 == p2) {
f2eace23 1507 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1da177e4
LT		 1508 return NULL;
1509 }
1510
a11f3a05 1511 mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex);
1da177e4
LT		 1512
1513 for (p = p1; p->d_parent != p; p = p->d_parent) {
1514 if (p->d_parent == p2) {
f2eace23
IM		 1515 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT);
1516 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD);
1da177e4
LT		 1517 return p;
1518 }
1519 }
1520
1521 for (p = p2; p->d_parent != p; p = p->d_parent) {
1522 if (p->d_parent == p1) {
f2eace23
IM		 1523 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1524 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD);
1da177e4
LT		 1525 return p;
1526 }
1527 }
1528
f2eace23
IM		 1529 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1530 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD);
1da177e4
LT		 1531 return NULL;
1532}
1533
1534void unlock_rename(struct dentry *p1, struct dentry *p2)
1535{
1b1dcc1b 1536 mutex_unlock(&p1->d_inode->i_mutex);
1da177e4 1537 if (p1 != p2) {
1b1dcc1b 1538 mutex_unlock(&p2->d_inode->i_mutex);
a11f3a05 1539 mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex);
1da177e4
LT		 1540 }
1541}
1542
1543int vfs_create(struct inode *dir, struct dentry *dentry, int mode,
1544 struct nameidata *nd)
1545{
1546 int error = may_create(dir, dentry, nd);
1547
1548 if (error)
1549 return error;
1550
1551 if (!dir->i_op || !dir->i_op->create)
1552 return -EACCES; /* shouldn't it be ENOSYS? */
1553 mode &= S_IALLUGO;
1554 mode |= S_IFREG;
1555 error = security_inode_create(dir, dentry, mode);
1556 if (error)
1557 return error;
1558 DQUOT_INIT(dir);
1559 error = dir->i_op->create(dir, dentry, mode, nd);
a74574aa 1560 if (!error)
f38aa942 1561 fsnotify_create(dir, dentry);
1da177e4
LT		 1562 return error;
1563}
1564
1565int may_open(struct nameidata *nd, int acc_mode, int flag)
1566{
1567 struct dentry *dentry = nd->dentry;
1568 struct inode *inode = dentry->d_inode;
1569 int error;
1570
1571 if (!inode)
1572 return -ENOENT;
1573
1574 if (S_ISLNK(inode->i_mode))
1575 return -ELOOP;
1576
1577 if (S_ISDIR(inode->i_mode) && (flag & FMODE_WRITE))
1578 return -EISDIR;
1579
e4543edd 1580 error = vfs_permission(nd, acc_mode);
1da177e4
LT		 1581 if (error)
1582 return error;
1583
1584 /*
1585 * FIFO's, sockets and device files are special: they don't
1586 * actually live on the filesystem itself, and as such you
1587 * can write to them even if the filesystem is read-only.
1588 */
1589 if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
1590 flag &= ~O_TRUNC;
1591 } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) {
1592 if (nd->mnt->mnt_flags & MNT_NODEV)
1593 return -EACCES;
1594
1595 flag &= ~O_TRUNC;
1596 } else if (IS_RDONLY(inode) && (flag & FMODE_WRITE))
1597 return -EROFS;
1598 /*
1599 * An append-only file must be opened in append mode for writing.
1600 */
1601 if (IS_APPEND(inode)) {
1602 if ((flag & FMODE_WRITE) && !(flag & O_APPEND))
1603 return -EPERM;
1604 if (flag & O_TRUNC)
1605 return -EPERM;
1606 }
1607
1608 /* O_NOATIME can only be set by the owner or superuser */
1609 if (flag & O_NOATIME)
3bd858ab 1610 if (!is_owner_or_cap(inode))
1da177e4
LT		 1611 return -EPERM;
1612
1613 /*
1614 * Ensure there are no outstanding leases on the file.
1615 */
1616 error = break_lease(inode, flag);
1617 if (error)
1618 return error;
1619
1620 if (flag & O_TRUNC) {
1621 error = get_write_access(inode);
1622 if (error)
1623 return error;
1624
1625 /*
1626 * Refuse to truncate files with mandatory locks held on them.
1627 */
1628 error = locks_verify_locked(inode);
1629 if (!error) {
1630 DQUOT_INIT(inode);
1631
4a30131e 1632 error = do_truncate(dentry, 0, ATTR_MTIME|ATTR_CTIME, NULL);
1da177e4
LT		 1633 }
1634 put_write_access(inode);
1635 if (error)
1636 return error;
1637 } else
1638 if (flag & FMODE_WRITE)
1639 DQUOT_INIT(inode);
1640
1641 return 0;
1642}
1643
aab520e2
DH		 1644static int open_namei_create(struct nameidata *nd, struct path *path,
1645 int flag, int mode)
1646{
1647 int error;
1648 struct dentry *dir = nd->dentry;
1649
1650 if (!IS_POSIXACL(dir->d_inode))
1651 mode &= ~current->fs->umask;
1652 error = vfs_create(dir->d_inode, path->dentry, mode, nd);
1653 mutex_unlock(&dir->d_inode->i_mutex);
1654 dput(nd->dentry);
1655 nd->dentry = path->dentry;
1656 if (error)
1657 return error;
1658 /* Don't check for write permission, don't truncate */
1659 return may_open(nd, 0, flag & ~O_TRUNC);
1660}
1661
1da177e4
LT		 1662/*
1663 * open_namei()
1664 *
1665 * namei for open - this is in fact almost the whole open-routine.
1666 *
1667 * Note that the low bits of "flag" aren't the same as in the open
1668 * system call - they are 00 - no permissions needed
1669 * 01 - read permission needed
1670 * 10 - write permission needed
1671 * 11 - read/write permissions needed
1672 * which is a lot more logical, and also allows the "no perm" needed
1673 * for symlinks (where the permissions are checked later).
1674 * SMP-safe
1675 */
5590ff0d
UD		 1676int open_namei(int dfd, const char *pathname, int flag,
1677 int mode, struct nameidata *nd)
1da177e4 1678{
834f2a4a 1679 int acc_mode, error;
4e7506e4 1680 struct path path;
1da177e4
LT		 1681 struct dentry *dir;
1682 int count = 0;
1683
1684 acc_mode = ACC_MODE(flag);
1685
834f2a4a
TM		 1686 /* O_TRUNC implies we need access checks for write permissions */
1687 if (flag & O_TRUNC)
1688 acc_mode |= MAY_WRITE;
1689
1da177e4
LT		 1690 /* Allow the LSM permission hook to distinguish append
1691 access from general write access. */
1692 if (flag & O_APPEND)
1693 acc_mode |= MAY_APPEND;
1694
1da177e4
LT		 1695 /*
1696 * The simplest case - just a plain lookup.
1697 */
1698 if (!(flag & O_CREAT)) {
5590ff0d
UD		 1699 error = path_lookup_open(dfd, pathname, lookup_flags(flag),
1700 nd, flag);
1da177e4
LT		 1701 if (error)
1702 return error;
1703 goto ok;
1704 }
1705
1706 /*
1707 * Create - we need to know the parent.
1708 */
5590ff0d 1709 error = path_lookup_create(dfd,pathname,LOOKUP_PARENT,nd,flag,mode);
1da177e4
LT		 1710 if (error)
1711 return error;
1712
1713 /*
1714 * We have the parent and last component. First of all, check
1715 * that we are not asked to creat(2) an obvious directory - that
1716 * will not do.
1717 */
1718 error = -EISDIR;
1719 if (nd->last_type != LAST_NORM || nd->last.name[nd->last.len])
1720 goto exit;
1721
1722 dir = nd->dentry;
1723 nd->flags &= ~LOOKUP_PARENT;
1b1dcc1b 1724 mutex_lock(&dir->d_inode->i_mutex);
49705b77 1725 path.dentry = lookup_hash(nd);
d73ffe16 1726 path.mnt = nd->mnt;
1da177e4
LT		 1727
1728do_last:
4e7506e4
AV		 1729 error = PTR_ERR(path.dentry);
1730 if (IS_ERR(path.dentry)) {
1b1dcc1b 1731 mutex_unlock(&dir->d_inode->i_mutex);
1da177e4
LT		 1732 goto exit;
1733 }
1734
4af4c52f
OD		 1735 if (IS_ERR(nd->intent.open.file)) {
1736 mutex_unlock(&dir->d_inode->i_mutex);
1737 error = PTR_ERR(nd->intent.open.file);
1738 goto exit_dput;
1739 }
1740
1da177e4 1741 /* Negative dentry, just create the file */
4e7506e4 1742 if (!path.dentry->d_inode) {
aab520e2 1743 error = open_namei_create(nd, &path, flag, mode);
1da177e4
LT		 1744 if (error)
1745 goto exit;
aab520e2 1746 return 0;
1da177e4
LT		 1747 }
1748
1749 /*
1750 * It already exists.
1751 */
1b1dcc1b 1752 mutex_unlock(&dir->d_inode->i_mutex);
4fc03b9b 1753 audit_inode(pathname, path.dentry->d_inode);
1da177e4
LT		 1754
1755 error = -EEXIST;
1756 if (flag & O_EXCL)
1757 goto exit_dput;
1758
e13b210f 1759 if (__follow_mount(&path)) {
1da177e4 1760 error = -ELOOP;
ba7a4c1a
AV		 1761 if (flag & O_NOFOLLOW)
1762 goto exit_dput;
1da177e4 1763 }
3e2efce0 1764
1da177e4 1765 error = -ENOENT;
4e7506e4 1766 if (!path.dentry->d_inode)
1da177e4 1767 goto exit_dput;
4e7506e4 1768 if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link)
1da177e4
LT		 1769 goto do_link;
1770
09dd17d3 1771 path_to_nameidata(&path, nd);
1da177e4 1772 error = -EISDIR;
4e7506e4 1773 if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode))
1da177e4
LT		 1774 goto exit;
1775ok:
1776 error = may_open(nd, acc_mode, flag);
1777 if (error)
1778 goto exit;
1779 return 0;
1780
1781exit_dput:
09dd17d3 1782 dput_path(&path, nd);
1da177e4 1783exit:
834f2a4a
TM		 1784 if (!IS_ERR(nd->intent.open.file))
1785 release_open_intent(nd);
1da177e4
LT		 1786 path_release(nd);
1787 return error;
1788
1789do_link:
1790 error = -ELOOP;
1791 if (flag & O_NOFOLLOW)
1792 goto exit_dput;
1793 /*
1794 * This is subtle. Instead of calling do_follow_link() we do the
1795 * thing by hands. The reason is that this way we have zero link_count
1796 * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT.
1797 * After that we have the parent and last component, i.e.
1798 * we are in the same situation as after the first path_walk().
1799 * Well, almost - if the last component is normal we get its copy
1800 * stored in nd->last.name and we will have to putname() it when we
1801 * are done. Procfs-like symlinks just set LAST_BIND.
1802 */
1803 nd->flags |= LOOKUP_PARENT;
4e7506e4 1804 error = security_inode_follow_link(path.dentry, nd);
1da177e4
LT		 1805 if (error)
1806 goto exit_dput;
cd4e91d3 1807 error = __do_follow_link(&path, nd);
de459215
KK		 1808 if (error) {
1809 /* Does someone understand code flow here? Or it is only
1810 * me so stupid? Anathema to whoever designed this non-sense
1811 * with "intent.open".
1812 */
1813 release_open_intent(nd);
1da177e4 1814 return error;
de459215 1815 }
1da177e4 1816 nd->flags &= ~LOOKUP_PARENT;
d671d5e5 1817 if (nd->last_type == LAST_BIND)
1da177e4 1818 goto ok;
1da177e4
LT		 1819 error = -EISDIR;
1820 if (nd->last_type != LAST_NORM)
1821 goto exit;
1822 if (nd->last.name[nd->last.len]) {
82984114 1823 __putname(nd->last.name);
1da177e4
LT		 1824 goto exit;
1825 }
1826 error = -ELOOP;
1827 if (count++==32) {
82984114 1828 __putname(nd->last.name);
1da177e4
LT		 1829 goto exit;
1830 }
1831 dir = nd->dentry;
1b1dcc1b 1832 mutex_lock(&dir->d_inode->i_mutex);
49705b77 1833 path.dentry = lookup_hash(nd);
d671d5e5 1834 path.mnt = nd->mnt;
82984114 1835 __putname(nd->last.name);
1da177e4
LT		 1836 goto do_last;
1837}
1838
1839/**
1840 * lookup_create - lookup a dentry, creating it if it doesn't exist
1841 * @nd: nameidata info
1842 * @is_dir: directory flag
1843 *
1844 * Simple function to lookup and return a dentry and create it
1845 * if it doesn't exist. Is SMP-safe.
c663e5d8 1846 *
1b1dcc1b 1847 * Returns with nd->dentry->d_inode->i_mutex locked.
1da177e4
LT		 1848 */
1849struct dentry *lookup_create(struct nameidata *nd, int is_dir)
1850{
c663e5d8 1851 struct dentry *dentry = ERR_PTR(-EEXIST);
1da177e4 1852
f2eace23 1853 mutex_lock_nested(&nd->dentry->d_inode->i_mutex, I_MUTEX_PARENT);
c663e5d8
CH		 1854 /*
1855 * Yucky last component or no last component at all?
1856 * (foo/., foo/.., /////)
1857 */
1da177e4
LT		 1858 if (nd->last_type != LAST_NORM)
1859 goto fail;
1860 nd->flags &= ~LOOKUP_PARENT;
a634904a
AM		 1861 nd->flags |= LOOKUP_CREATE;
1862 nd->intent.open.flags = O_EXCL;
c663e5d8
CH		 1863
1864 /*
1865 * Do the final lookup.
1866 */
49705b77 1867 dentry = lookup_hash(nd);
1da177e4
LT		 1868 if (IS_ERR(dentry))
1869 goto fail;
c663e5d8
CH		 1870
1871 /*
1872 * Special case - lookup gave negative, but... we had foo/bar/
1873 * From the vfs_mknod() POV we just have a negative dentry -
1874 * all is fine. Let's be bastards - you had / on the end, you've
1875 * been asking for (non-existent) directory. -ENOENT for you.
1876 */
1da177e4
LT		 1877 if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
1878 goto enoent;
1879 return dentry;
1880enoent:
1881 dput(dentry);
1882 dentry = ERR_PTR(-ENOENT);
1883fail:
1884 return dentry;
1885}
f81a0bff 1886EXPORT_SYMBOL_GPL(lookup_create);
1da177e4
LT		 1887
1888int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
1889{
1890 int error = may_create(dir, dentry, NULL);
1891
1892 if (error)
1893 return error;
1894
1895 if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD))
1896 return -EPERM;
1897
1898 if (!dir->i_op || !dir->i_op->mknod)
1899 return -EPERM;
1900
1901 error = security_inode_mknod(dir, dentry, mode, dev);
1902 if (error)
1903 return error;
1904
1905 DQUOT_INIT(dir);
1906 error = dir->i_op->mknod(dir, dentry, mode, dev);
a74574aa 1907 if (!error)
f38aa942 1908 fsnotify_create(dir, dentry);
1da177e4
LT		 1909 return error;
1910}
1911
5590ff0d
UD		 1912asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode,
1913 unsigned dev)
1da177e4
LT		 1914{
1915 int error = 0;
1916 char * tmp;
1917 struct dentry * dentry;
1918 struct nameidata nd;
1919
1920 if (S_ISDIR(mode))
1921 return -EPERM;
1922 tmp = getname(filename);
1923 if (IS_ERR(tmp))
1924 return PTR_ERR(tmp);
1925
5590ff0d 1926 error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd);
1da177e4
LT		 1927 if (error)
1928 goto out;
1929 dentry = lookup_create(&nd, 0);
1930 error = PTR_ERR(dentry);
1931
1932 if (!IS_POSIXACL(nd.dentry->d_inode))
1933 mode &= ~current->fs->umask;
1934 if (!IS_ERR(dentry)) {
1935 switch (mode & S_IFMT) {
1936 case 0: case S_IFREG:
1937 error = vfs_create(nd.dentry->d_inode,dentry,mode,&nd);
1938 break;
1939 case S_IFCHR: case S_IFBLK:
1940 error = vfs_mknod(nd.dentry->d_inode,dentry,mode,
1941 new_decode_dev(dev));
1942 break;
1943 case S_IFIFO: case S_IFSOCK:
1944 error = vfs_mknod(nd.dentry->d_inode,dentry,mode,0);
1945 break;
1946 case S_IFDIR:
1947 error = -EPERM;
1948 break;
1949 default:
1950 error = -EINVAL;
1951 }
1952 dput(dentry);
1953 }
1b1dcc1b 1954 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 1955 path_release(&nd);
1956out:
1957 putname(tmp);
1958
1959 return error;
1960}
1961
5590ff0d
UD		 1962asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev)
1963{
1964 return sys_mknodat(AT_FDCWD, filename, mode, dev);
1965}
1966
1da177e4
LT		 1967int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1968{
1969 int error = may_create(dir, dentry, NULL);
1970
1971 if (error)
1972 return error;
1973
1974 if (!dir->i_op || !dir->i_op->mkdir)
1975 return -EPERM;
1976
1977 mode &= (S_IRWXUGO|S_ISVTX);
1978 error = security_inode_mkdir(dir, dentry, mode);
1979 if (error)
1980 return error;
1981
1982 DQUOT_INIT(dir);
1983 error = dir->i_op->mkdir(dir, dentry, mode);
a74574aa 1984 if (!error)
f38aa942 1985 fsnotify_mkdir(dir, dentry);
1da177e4
LT		 1986 return error;
1987}
1988
5590ff0d 1989asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode)
1da177e4
LT		 1990{
1991 int error = 0;
1992 char * tmp;
6902d925
DH		 1993 struct dentry *dentry;
1994 struct nameidata nd;
1da177e4
LT		 1995
1996 tmp = getname(pathname);
1997 error = PTR_ERR(tmp);
6902d925
DH		 1998 if (IS_ERR(tmp))
1999 goto out_err;
1da177e4 2000
6902d925
DH		 2001 error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd);
2002 if (error)
2003 goto out;
2004 dentry = lookup_create(&nd, 1);
2005 error = PTR_ERR(dentry);
2006 if (IS_ERR(dentry))
2007 goto out_unlock;
1da177e4 2008
6902d925
DH		 2009 if (!IS_POSIXACL(nd.dentry->d_inode))
2010 mode &= ~current->fs->umask;
2011 error = vfs_mkdir(nd.dentry->d_inode, dentry, mode);
2012 dput(dentry);
2013out_unlock:
2014 mutex_unlock(&nd.dentry->d_inode->i_mutex);
2015 path_release(&nd);
2016out:
2017 putname(tmp);
2018out_err:
1da177e4
LT		 2019 return error;
2020}
2021
5590ff0d
UD		 2022asmlinkage long sys_mkdir(const char __user *pathname, int mode)
2023{
2024 return sys_mkdirat(AT_FDCWD, pathname, mode);
2025}
2026
1da177e4
LT		 2027/*
2028 * We try to drop the dentry early: we should have
2029 * a usage count of 2 if we're the only user of this
2030 * dentry, and if that is true (possibly after pruning
2031 * the dcache), then we drop the dentry now.
2032 *
2033 * A low-level filesystem can, if it choses, legally
2034 * do a
2035 *
2036 * if (!d_unhashed(dentry))
2037 * return -EBUSY;
2038 *
2039 * if it cannot handle the case of removing a directory
2040 * that is still in use by something else..
2041 */
2042void dentry_unhash(struct dentry *dentry)
2043{
2044 dget(dentry);
dc168427 2045 shrink_dcache_parent(dentry);
1da177e4
LT		 2046 spin_lock(&dcache_lock);
2047 spin_lock(&dentry->d_lock);
2048 if (atomic_read(&dentry->d_count) == 2)
2049 __d_drop(dentry);
2050 spin_unlock(&dentry->d_lock);
2051 spin_unlock(&dcache_lock);
2052}
2053
2054int vfs_rmdir(struct inode *dir, struct dentry *dentry)
2055{
2056 int error = may_delete(dir, dentry, 1);
2057
2058 if (error)
2059 return error;
2060
2061 if (!dir->i_op || !dir->i_op->rmdir)
2062 return -EPERM;
2063
2064 DQUOT_INIT(dir);
2065
1b1dcc1b 2066 mutex_lock(&dentry->d_inode->i_mutex);
1da177e4
LT		 2067 dentry_unhash(dentry);
2068 if (d_mountpoint(dentry))
2069 error = -EBUSY;
2070 else {
2071 error = security_inode_rmdir(dir, dentry);
2072 if (!error) {
2073 error = dir->i_op->rmdir(dir, dentry);
2074 if (!error)
2075 dentry->d_inode->i_flags |= S_DEAD;
2076 }
2077 }
1b1dcc1b 2078 mutex_unlock(&dentry->d_inode->i_mutex);
1da177e4 2079 if (!error) {
1da177e4
LT		 2080 d_delete(dentry);
2081 }
2082 dput(dentry);
2083
2084 return error;
2085}
2086
5590ff0d 2087static long do_rmdir(int dfd, const char __user *pathname)
1da177e4
LT		 2088{
2089 int error = 0;
2090 char * name;
2091 struct dentry *dentry;
2092 struct nameidata nd;
2093
2094 name = getname(pathname);
2095 if(IS_ERR(name))
2096 return PTR_ERR(name);
2097
5590ff0d 2098 error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd);
1da177e4
LT		 2099 if (error)
2100 goto exit;
2101
2102 switch(nd.last_type) {
2103 case LAST_DOTDOT:
2104 error = -ENOTEMPTY;
2105 goto exit1;
2106 case LAST_DOT:
2107 error = -EINVAL;
2108 goto exit1;
2109 case LAST_ROOT:
2110 error = -EBUSY;
2111 goto exit1;
2112 }
f2eace23 2113 mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT);
49705b77 2114 dentry = lookup_hash(&nd);
1da177e4 2115 error = PTR_ERR(dentry);
6902d925
DH		 2116 if (IS_ERR(dentry))
2117 goto exit2;
2118 error = vfs_rmdir(nd.dentry->d_inode, dentry);
2119 dput(dentry);
2120exit2:
1b1dcc1b 2121 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2122exit1:
2123 path_release(&nd);
2124exit:
2125 putname(name);
2126 return error;
2127}
2128
5590ff0d
UD		 2129asmlinkage long sys_rmdir(const char __user *pathname)
2130{
2131 return do_rmdir(AT_FDCWD, pathname);
2132}
2133
1da177e4
LT		 2134int vfs_unlink(struct inode *dir, struct dentry *dentry)
2135{
2136 int error = may_delete(dir, dentry, 0);
2137
2138 if (error)
2139 return error;
2140
2141 if (!dir->i_op || !dir->i_op->unlink)
2142 return -EPERM;
2143
2144 DQUOT_INIT(dir);
2145
1b1dcc1b 2146 mutex_lock(&dentry->d_inode->i_mutex);
1da177e4
LT		 2147 if (d_mountpoint(dentry))
2148 error = -EBUSY;
2149 else {
2150 error = security_inode_unlink(dir, dentry);
2151 if (!error)
2152 error = dir->i_op->unlink(dir, dentry);
2153 }
1b1dcc1b 2154 mutex_unlock(&dentry->d_inode->i_mutex);
1da177e4
LT		 2155
2156 /* We don't d_delete() NFS sillyrenamed files--they still exist. */
2157 if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) {
e234f35c 2158 d_delete(dentry);
1da177e4 2159 }
0eeca283 2160
1da177e4
LT		 2161 return error;
2162}
2163
2164/*
2165 * Make sure that the actual truncation of the file will occur outside its
1b1dcc1b 2166 * directory's i_mutex. Truncate can take a long time if there is a lot of
1da177e4
LT		 2167 * writeout happening, and we don't want to prevent access to the directory
2168 * while waiting on the I/O.
2169 */
5590ff0d 2170static long do_unlinkat(int dfd, const char __user *pathname)
1da177e4
LT		 2171{
2172 int error = 0;
2173 char * name;
2174 struct dentry *dentry;
2175 struct nameidata nd;
2176 struct inode *inode = NULL;
2177
2178 name = getname(pathname);
2179 if(IS_ERR(name))
2180 return PTR_ERR(name);
2181
5590ff0d 2182 error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd);
1da177e4
LT		 2183 if (error)
2184 goto exit;
2185 error = -EISDIR;
2186 if (nd.last_type != LAST_NORM)
2187 goto exit1;
f2eace23 2188 mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT);
49705b77 2189 dentry = lookup_hash(&nd);
1da177e4
LT		 2190 error = PTR_ERR(dentry);
2191 if (!IS_ERR(dentry)) {
2192 /* Why not before? Because we want correct error value */
2193 if (nd.last.name[nd.last.len])
2194 goto slashes;
2195 inode = dentry->d_inode;
2196 if (inode)
2197 atomic_inc(&inode->i_count);
2198 error = vfs_unlink(nd.dentry->d_inode, dentry);
2199 exit2:
2200 dput(dentry);
2201 }
1b1dcc1b 2202 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2203 if (inode)
2204 iput(inode); /* truncate the inode here */
2205exit1:
2206 path_release(&nd);
2207exit:
2208 putname(name);
2209 return error;
2210
2211slashes:
2212 error = !dentry->d_inode ? -ENOENT :
2213 S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR;
2214 goto exit2;
2215}
2216
5590ff0d
UD		 2217asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag)
2218{
2219 if ((flag & ~AT_REMOVEDIR) != 0)
2220 return -EINVAL;
2221
2222 if (flag & AT_REMOVEDIR)
2223 return do_rmdir(dfd, pathname);
2224
2225 return do_unlinkat(dfd, pathname);
2226}
2227
2228asmlinkage long sys_unlink(const char __user *pathname)
2229{
2230 return do_unlinkat(AT_FDCWD, pathname);
2231}
2232
1da177e4
LT		 2233int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname, int mode)
2234{
2235 int error = may_create(dir, dentry, NULL);
2236
2237 if (error)
2238 return error;
2239
2240 if (!dir->i_op || !dir->i_op->symlink)
2241 return -EPERM;
2242
2243 error = security_inode_symlink(dir, dentry, oldname);
2244 if (error)
2245 return error;
2246
2247 DQUOT_INIT(dir);
2248 error = dir->i_op->symlink(dir, dentry, oldname);
a74574aa 2249 if (!error)
f38aa942 2250 fsnotify_create(dir, dentry);
1da177e4
LT		 2251 return error;
2252}
2253
5590ff0d
UD		 2254asmlinkage long sys_symlinkat(const char __user *oldname,
2255 int newdfd, const char __user *newname)
1da177e4
LT		 2256{
2257 int error = 0;
2258 char * from;
2259 char * to;
6902d925
DH		 2260 struct dentry *dentry;
2261 struct nameidata nd;
1da177e4
LT		 2262
2263 from = getname(oldname);
2264 if(IS_ERR(from))
2265 return PTR_ERR(from);
2266 to = getname(newname);
2267 error = PTR_ERR(to);
6902d925
DH		 2268 if (IS_ERR(to))
2269 goto out_putname;
1da177e4 2270
6902d925
DH		 2271 error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd);
2272 if (error)
2273 goto out;
2274 dentry = lookup_create(&nd, 0);
2275 error = PTR_ERR(dentry);
2276 if (IS_ERR(dentry))
2277 goto out_unlock;
2278
2279 error = vfs_symlink(nd.dentry->d_inode, dentry, from, S_IALLUGO);
2280 dput(dentry);
2281out_unlock:
2282 mutex_unlock(&nd.dentry->d_inode->i_mutex);
2283 path_release(&nd);
1da177e4 2284out:
6902d925
DH		 2285 putname(to);
2286out_putname:
1da177e4
LT		 2287 putname(from);
2288 return error;
2289}
2290
5590ff0d
UD		 2291asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname)
2292{
2293 return sys_symlinkat(oldname, AT_FDCWD, newname);
2294}
2295
1da177e4
LT		 2296int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry)
2297{
2298 struct inode *inode = old_dentry->d_inode;
2299 int error;
2300
2301 if (!inode)
2302 return -ENOENT;
2303
2304 error = may_create(dir, new_dentry, NULL);
2305 if (error)
2306 return error;
2307
2308 if (dir->i_sb != inode->i_sb)
2309 return -EXDEV;
2310
2311 /*
2312 * A link to an append-only or immutable file cannot be created.
2313 */
2314 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2315 return -EPERM;
2316 if (!dir->i_op || !dir->i_op->link)
2317 return -EPERM;
2318 if (S_ISDIR(old_dentry->d_inode->i_mode))
2319 return -EPERM;
2320
2321 error = security_inode_link(old_dentry, dir, new_dentry);
2322 if (error)
2323 return error;
2324
1b1dcc1b 2325 mutex_lock(&old_dentry->d_inode->i_mutex);
1da177e4
LT		 2326 DQUOT_INIT(dir);
2327 error = dir->i_op->link(old_dentry, dir, new_dentry);
1b1dcc1b 2328 mutex_unlock(&old_dentry->d_inode->i_mutex);
e31e14ec 2329 if (!error)
f38aa942 2330 fsnotify_create(dir, new_dentry);
1da177e4
LT		 2331 return error;
2332}
2333
2334/*
2335 * Hardlinks are often used in delicate situations. We avoid
2336 * security-related surprises by not following symlinks on the
2337 * newname. --KAB
2338 *
2339 * We don't follow them on the oldname either to be compatible
2340 * with linux 2.0, and to avoid hard-linking to directories
2341 * and other special files. --ADM
2342 */
5590ff0d 2343asmlinkage long sys_linkat(int olddfd, const char __user *oldname,
c04030e1
UD		 2344 int newdfd, const char __user *newname,
2345 int flags)
1da177e4
LT		 2346{
2347 struct dentry *new_dentry;
2348 struct nameidata nd, old_nd;
2349 int error;
2350 char * to;
2351
45c9b11a 2352 if ((flags & ~AT_SYMLINK_FOLLOW) != 0)
c04030e1
UD		 2353 return -EINVAL;
2354
1da177e4
LT		 2355 to = getname(newname);
2356 if (IS_ERR(to))
2357 return PTR_ERR(to);
2358
45c9b11a
UD		 2359 error = __user_walk_fd(olddfd, oldname,
2360 flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0,
2361 &old_nd);
1da177e4
LT		 2362 if (error)
2363 goto exit;
5590ff0d 2364 error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd);
1da177e4
LT		 2365 if (error)
2366 goto out;
2367 error = -EXDEV;
2368 if (old_nd.mnt != nd.mnt)
2369 goto out_release;
2370 new_dentry = lookup_create(&nd, 0);
2371 error = PTR_ERR(new_dentry);
6902d925
DH		 2372 if (IS_ERR(new_dentry))
2373 goto out_unlock;
2374 error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry);
2375 dput(new_dentry);
2376out_unlock:
1b1dcc1b 2377 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2378out_release:
2379 path_release(&nd);
2380out:
2381 path_release(&old_nd);
2382exit:
2383 putname(to);
2384
2385 return error;
2386}
2387
5590ff0d
UD		 2388asmlinkage long sys_link(const char __user *oldname, const char __user *newname)
2389{
c04030e1 2390 return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0);
5590ff0d
UD		 2391}
2392
1da177e4
LT		 2393/*
2394 * The worst of all namespace operations - renaming directory. "Perverted"
2395 * doesn't even start to describe it. Somebody in UCB had a heck of a trip...
2396 * Problems:
2397 * a) we can get into loop creation. Check is done in is_subdir().
2398 * b) race potential - two innocent renames can create a loop together.
2399 * That's where 4.4 screws up. Current fix: serialization on
a11f3a05 2400 * sb->s_vfs_rename_mutex. We might be more accurate, but that's another
1da177e4
LT		 2401 * story.
2402 * c) we have to lock _three_ objects - parents and victim (if it exists).
1b1dcc1b 2403 * And that - after we got ->i_mutex on parents (until then we don't know
1da177e4
LT		 2404 * whether the target exists). Solution: try to be smart with locking
2405 * order for inodes. We rely on the fact that tree topology may change
a11f3a05 2406 * only under ->s_vfs_rename_mutex _and_ that parent of the object we
1da177e4
LT		 2407 * move will be locked. Thus we can rank directories by the tree
2408 * (ancestors first) and rank all non-directories after them.
2409 * That works since everybody except rename does "lock parent, lookup,
a11f3a05 2410 * lock child" and rename is under ->s_vfs_rename_mutex.
1da177e4
LT		 2411 * HOWEVER, it relies on the assumption that any object with ->lookup()
2412 * has no more than 1 dentry. If "hybrid" objects will ever appear,
2413 * we'd better make sure that there's no link(2) for them.
2414 * d) some filesystems don't support opened-but-unlinked directories,
2415 * either because of layout or because they are not ready to deal with
2416 * all cases correctly. The latter will be fixed (taking this sort of
2417 * stuff into VFS), but the former is not going away. Solution: the same
2418 * trick as in rmdir().
2419 * e) conversion from fhandle to dentry may come in the wrong moment - when
1b1dcc1b 2420 * we are removing the target. Solution: we will have to grab ->i_mutex
1da177e4 2421 * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on
1b1dcc1b 2422 * ->i_mutex on parents, which works but leads to some truely excessive
1da177e4
LT		 2423 * locking].
2424 */
75c96f85
AB		 2425static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry,
2426 struct inode *new_dir, struct dentry *new_dentry)
1da177e4
LT		 2427{
2428 int error = 0;
2429 struct inode *target;
2430
2431 /*
2432 * If we are going to change the parent - check write permissions,
2433 * we'll need to flip '..'.
2434 */
2435 if (new_dir != old_dir) {
2436 error = permission(old_dentry->d_inode, MAY_WRITE, NULL);
2437 if (error)
2438 return error;
2439 }
2440
2441 error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry);
2442 if (error)
2443 return error;
2444
2445 target = new_dentry->d_inode;
2446 if (target) {
1b1dcc1b 2447 mutex_lock(&target->i_mutex);
1da177e4
LT		 2448 dentry_unhash(new_dentry);
2449 }
2450 if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry))
2451 error = -EBUSY;
2452 else
2453 error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry);
2454 if (target) {
2455 if (!error)
2456 target->i_flags |= S_DEAD;
1b1dcc1b 2457 mutex_unlock(&target->i_mutex);
1da177e4
LT		 2458 if (d_unhashed(new_dentry))
2459 d_rehash(new_dentry);
2460 dput(new_dentry);
2461 }
e31e14ec 2462 if (!error)
349457cc
MF		 2463 if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE))
2464 d_move(old_dentry,new_dentry);
1da177e4
LT		 2465 return error;
2466}
2467
75c96f85
AB		 2468static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry,
2469 struct inode *new_dir, struct dentry *new_dentry)
1da177e4
LT		 2470{
2471 struct inode *target;
2472 int error;
2473
2474 error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry);
2475 if (error)
2476 return error;
2477
2478 dget(new_dentry);
2479 target = new_dentry->d_inode;
2480 if (target)
1b1dcc1b 2481 mutex_lock(&target->i_mutex);
1da177e4
LT		 2482 if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry))
2483 error = -EBUSY;
2484 else
2485 error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry);
2486 if (!error) {
349457cc 2487 if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE))
1da177e4 2488 d_move(old_dentry, new_dentry);
1da177e4
LT		 2489 }
2490 if (target)
1b1dcc1b 2491 mutex_unlock(&target->i_mutex);
1da177e4
LT		 2492 dput(new_dentry);
2493 return error;
2494}
2495
2496int vfs_rename(struct inode *old_dir, struct dentry *old_dentry,
2497 struct inode *new_dir, struct dentry *new_dentry)
2498{
2499 int error;
2500 int is_dir = S_ISDIR(old_dentry->d_inode->i_mode);
0eeca283 2501 const char *old_name;
1da177e4
LT		 2502
2503 if (old_dentry->d_inode == new_dentry->d_inode)
2504 return 0;
2505
2506 error = may_delete(old_dir, old_dentry, is_dir);
2507 if (error)
2508 return error;
2509
2510 if (!new_dentry->d_inode)
2511 error = may_create(new_dir, new_dentry, NULL);
2512 else
2513 error = may_delete(new_dir, new_dentry, is_dir);
2514 if (error)
2515 return error;
2516
2517 if (!old_dir->i_op || !old_dir->i_op->rename)
2518 return -EPERM;
2519
2520 DQUOT_INIT(old_dir);
2521 DQUOT_INIT(new_dir);
2522
0eeca283
RL		 2523 old_name = fsnotify_oldname_init(old_dentry->d_name.name);
2524
1da177e4
LT		 2525 if (is_dir)
2526 error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry);
2527 else
2528 error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry);
2529 if (!error) {
0eeca283 2530 const char *new_name = old_dentry->d_name.name;
89204c40
JM		 2531 fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir,
2532 new_dentry->d_inode, old_dentry->d_inode);
1da177e4 2533 }
0eeca283
RL		 2534 fsnotify_oldname_free(old_name);
2535
1da177e4
LT		 2536 return error;
2537}
2538
5590ff0d
UD		 2539static int do_rename(int olddfd, const char *oldname,
2540 int newdfd, const char *newname)
1da177e4
LT		 2541{
2542 int error = 0;
2543 struct dentry * old_dir, * new_dir;
2544 struct dentry * old_dentry, *new_dentry;
2545 struct dentry * trap;
2546 struct nameidata oldnd, newnd;
2547
5590ff0d 2548 error = do_path_lookup(olddfd, oldname, LOOKUP_PARENT, &oldnd);
1da177e4
LT		 2549 if (error)
2550 goto exit;
2551
5590ff0d 2552 error = do_path_lookup(newdfd, newname, LOOKUP_PARENT, &newnd);
1da177e4
LT		 2553 if (error)
2554 goto exit1;
2555
2556 error = -EXDEV;
2557 if (oldnd.mnt != newnd.mnt)
2558 goto exit2;
2559
2560 old_dir = oldnd.dentry;
2561 error = -EBUSY;
2562 if (oldnd.last_type != LAST_NORM)
2563 goto exit2;
2564
2565 new_dir = newnd.dentry;
2566 if (newnd.last_type != LAST_NORM)
2567 goto exit2;
2568
2569 trap = lock_rename(new_dir, old_dir);
2570
49705b77 2571 old_dentry = lookup_hash(&oldnd);
1da177e4
LT		 2572 error = PTR_ERR(old_dentry);
2573 if (IS_ERR(old_dentry))
2574 goto exit3;
2575 /* source must exist */
2576 error = -ENOENT;
2577 if (!old_dentry->d_inode)
2578 goto exit4;
2579 /* unless the source is a directory trailing slashes give -ENOTDIR */
2580 if (!S_ISDIR(old_dentry->d_inode->i_mode)) {
2581 error = -ENOTDIR;
2582 if (oldnd.last.name[oldnd.last.len])
2583 goto exit4;
2584 if (newnd.last.name[newnd.last.len])
2585 goto exit4;
2586 }
2587 /* source should not be ancestor of target */
2588 error = -EINVAL;
2589 if (old_dentry == trap)
2590 goto exit4;
49705b77 2591 new_dentry = lookup_hash(&newnd);
1da177e4
LT		 2592 error = PTR_ERR(new_dentry);
2593 if (IS_ERR(new_dentry))
2594 goto exit4;
2595 /* target should not be an ancestor of source */
2596 error = -ENOTEMPTY;
2597 if (new_dentry == trap)
2598 goto exit5;
2599
2600 error = vfs_rename(old_dir->d_inode, old_dentry,
2601 new_dir->d_inode, new_dentry);
2602exit5:
2603 dput(new_dentry);
2604exit4:
2605 dput(old_dentry);
2606exit3:
2607 unlock_rename(new_dir, old_dir);
2608exit2:
2609 path_release(&newnd);
2610exit1:
2611 path_release(&oldnd);
2612exit:
2613 return error;
2614}
2615
5590ff0d
UD		 2616asmlinkage long sys_renameat(int olddfd, const char __user *oldname,
2617 int newdfd, const char __user *newname)
1da177e4
LT		 2618{
2619 int error;
2620 char * from;
2621 char * to;
2622
2623 from = getname(oldname);
2624 if(IS_ERR(from))
2625 return PTR_ERR(from);
2626 to = getname(newname);
2627 error = PTR_ERR(to);
2628 if (!IS_ERR(to)) {
5590ff0d 2629 error = do_rename(olddfd, from, newdfd, to);
1da177e4
LT		 2630 putname(to);
2631 }
2632 putname(from);
2633 return error;
2634}
2635
5590ff0d
UD		 2636asmlinkage long sys_rename(const char __user *oldname, const char __user *newname)
2637{
2638 return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname);
2639}
2640
1da177e4
LT		 2641int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link)
2642{
2643 int len;
2644
2645 len = PTR_ERR(link);
2646 if (IS_ERR(link))
2647 goto out;
2648
2649 len = strlen(link);
2650 if (len > (unsigned) buflen)
2651 len = buflen;
2652 if (copy_to_user(buffer, link, len))
2653 len = -EFAULT;
2654out:
2655 return len;
2656}
2657
2658/*
2659 * A helper for ->readlink(). This should be used *ONLY* for symlinks that
2660 * have ->follow_link() touching nd only in nd_set_link(). Using (or not
2661 * using) it for any given inode is up to filesystem.
2662 */
2663int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen)
2664{
2665 struct nameidata nd;
cc314eef
LT		 2666 void *cookie;
2667
1da177e4 2668 nd.depth = 0;
cc314eef
LT		 2669 cookie = dentry->d_inode->i_op->follow_link(dentry, &nd);
2670 if (!IS_ERR(cookie)) {
2671 int res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd));
1da177e4 2672 if (dentry->d_inode->i_op->put_link)
cc314eef
LT		 2673 dentry->d_inode->i_op->put_link(dentry, &nd, cookie);
2674 cookie = ERR_PTR(res);
1da177e4 2675 }
cc314eef 2676 return PTR_ERR(cookie);
1da177e4
LT		 2677}
2678
2679int vfs_follow_link(struct nameidata *nd, const char *link)
2680{
2681 return __vfs_follow_link(nd, link);
2682}
2683
2684/* get the link contents into pagecache */
2685static char *page_getlink(struct dentry * dentry, struct page **ppage)
2686{
2687 struct page * page;
2688 struct address_space *mapping = dentry->d_inode->i_mapping;
090d2b18 2689 page = read_mapping_page(mapping, 0, NULL);
1da177e4 2690 if (IS_ERR(page))
6fe6900e 2691 return (char*)page;
1da177e4
LT		 2692 *ppage = page;
2693 return kmap(page);
1da177e4
LT		 2694}
2695
2696int page_readlink(struct dentry *dentry, char __user *buffer, int buflen)
2697{
2698 struct page *page = NULL;
2699 char *s = page_getlink(dentry, &page);
2700 int res = vfs_readlink(dentry,buffer,buflen,s);
2701 if (page) {
2702 kunmap(page);
2703 page_cache_release(page);
2704 }
2705 return res;
2706}
2707
cc314eef 2708void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd)
1da177e4 2709{
cc314eef 2710 struct page *page = NULL;
1da177e4 2711 nd_set_link(nd, page_getlink(dentry, &page));
cc314eef 2712 return page;
1da177e4
LT		 2713}
2714
cc314eef 2715void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie)
1da177e4 2716{
cc314eef
LT		 2717 struct page *page = cookie;
2718
2719 if (page) {
1da177e4
LT		 2720 kunmap(page);
2721 page_cache_release(page);
1da177e4
LT		 2722 }
2723}
2724
0adb25d2
KK		 2725int __page_symlink(struct inode *inode, const char *symname, int len,
2726 gfp_t gfp_mask)
1da177e4
LT		 2727{
2728 struct address_space *mapping = inode->i_mapping;
0adb25d2 2729 struct page *page;
beb497ab 2730 int err;
1da177e4
LT		 2731 char *kaddr;
2732
7e53cac4 2733retry:
beb497ab 2734 err = -ENOMEM;
0adb25d2 2735 page = find_or_create_page(mapping, 0, gfp_mask);
1da177e4
LT		 2736 if (!page)
2737 goto fail;
2738 err = mapping->a_ops->prepare_write(NULL, page, 0, len-1);
7e53cac4
N		 2739 if (err == AOP_TRUNCATED_PAGE) {
2740 page_cache_release(page);
2741 goto retry;
2742 }
1da177e4
LT		 2743 if (err)
2744 goto fail_map;
2745 kaddr = kmap_atomic(page, KM_USER0);
2746 memcpy(kaddr, symname, len-1);
2747 kunmap_atomic(kaddr, KM_USER0);
7e53cac4
N		 2748 err = mapping->a_ops->commit_write(NULL, page, 0, len-1);
2749 if (err == AOP_TRUNCATED_PAGE) {
2750 page_cache_release(page);
2751 goto retry;
2752 }
2753 if (err)
2754 goto fail_map;
1da177e4
LT		 2755 /*
2756 * Notice that we are _not_ going to block here - end of page is
2757 * unmapped, so this will only try to map the rest of page, see
2758 * that it is unmapped (typically even will not look into inode -
2759 * ->i_size will be enough for everything) and zero it out.
2760 * OTOH it's obviously correct and should make the page up-to-date.
2761 */
2762 if (!PageUptodate(page)) {
2763 err = mapping->a_ops->readpage(NULL, page);
7e53cac4
N		 2764 if (err != AOP_TRUNCATED_PAGE)
2765 wait_on_page_locked(page);
1da177e4
LT		 2766 } else {
2767 unlock_page(page);
2768 }
2769 page_cache_release(page);
2770 if (err < 0)
2771 goto fail;
2772 mark_inode_dirty(inode);
2773 return 0;
2774fail_map:
2775 unlock_page(page);
2776 page_cache_release(page);
2777fail:
2778 return err;
2779}
2780
0adb25d2
KK		 2781int page_symlink(struct inode *inode, const char *symname, int len)
2782{
2783 return __page_symlink(inode, symname, len,
2784 mapping_gfp_mask(inode->i_mapping));
2785}
2786
92e1d5be 2787const struct inode_operations page_symlink_inode_operations = {
1da177e4
LT		 2788 .readlink = generic_readlink,
2789 .follow_link = page_follow_link_light,
2790 .put_link = page_put_link,
2791};
2792
2793EXPORT_SYMBOL(__user_walk);
5590ff0d 2794EXPORT_SYMBOL(__user_walk_fd);
1da177e4
LT		 2795EXPORT_SYMBOL(follow_down);
2796EXPORT_SYMBOL(follow_up);
2797EXPORT_SYMBOL(get_write_access); /* binfmt_aout */
2798EXPORT_SYMBOL(getname);
2799EXPORT_SYMBOL(lock_rename);
1da177e4
LT		 2800EXPORT_SYMBOL(lookup_one_len);
2801EXPORT_SYMBOL(page_follow_link_light);
2802EXPORT_SYMBOL(page_put_link);
2803EXPORT_SYMBOL(page_readlink);
0adb25d2 2804EXPORT_SYMBOL(__page_symlink);
1da177e4
LT		 2805EXPORT_SYMBOL(page_symlink);
2806EXPORT_SYMBOL(page_symlink_inode_operations);
2807EXPORT_SYMBOL(path_lookup);
16f18200 2808EXPORT_SYMBOL(vfs_path_lookup);
1da177e4
LT		 2809EXPORT_SYMBOL(path_release);
2810EXPORT_SYMBOL(path_walk);
2811EXPORT_SYMBOL(permission);
e4543edd 2812EXPORT_SYMBOL(vfs_permission);
8c744fb8 2813EXPORT_SYMBOL(file_permission);
1da177e4
LT		 2814EXPORT_SYMBOL(unlock_rename);
2815EXPORT_SYMBOL(vfs_create);
2816EXPORT_SYMBOL(vfs_follow_link);
2817EXPORT_SYMBOL(vfs_link);
2818EXPORT_SYMBOL(vfs_mkdir);
2819EXPORT_SYMBOL(vfs_mknod);
2820EXPORT_SYMBOL(generic_permission);
2821EXPORT_SYMBOL(vfs_readlink);
2822EXPORT_SYMBOL(vfs_rename);
2823EXPORT_SYMBOL(vfs_rmdir);
2824EXPORT_SYMBOL(vfs_symlink);
2825EXPORT_SYMBOL(vfs_unlink);
2826EXPORT_SYMBOL(dentry_unhash);
2827EXPORT_SYMBOL(generic_readlink);
Linux kernel - Deliverables
This page took 0.72605 seconds and 5 git commands to generate.