projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[PATCH] mark context of syscall entered with no rules as dummy
[GitHub/mt8127/android_kernel_alcatel_ttab.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/smp_lock.h>
26#include <linux/personality.h>
27#include <linux/security.h>
28#include <linux/syscalls.h>
29#include <linux/mount.h>
30#include <linux/audit.h>
16f7e0fe 31#include <linux/capability.h>
834f2a4a 32#include <linux/file.h>
5590ff0d
UD		 33#include <linux/fcntl.h>
34#include <linux/namei.h>
1da177e4
LT		 35#include <asm/namei.h>
36#include <asm/uaccess.h>
37
38#define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE])
39
40/* [Feb-1997 T. Schoebel-Theuer]
41 * Fundamental changes in the pathname lookup mechanisms (namei)
42 * were necessary because of omirr. The reason is that omirr needs
43 * to know the _real_ pathname, not the user-supplied one, in case
44 * of symlinks (and also when transname replacements occur).
45 *
46 * The new code replaces the old recursive symlink resolution with
47 * an iterative one (in case of non-nested symlink chains). It does
48 * this with calls to <fs>_follow_link().
49 * As a side effect, dir_namei(), _namei() and follow_link() are now
50 * replaced with a single function lookup_dentry() that can handle all
51 * the special cases of the former code.
52 *
53 * With the new dcache, the pathname is stored at each inode, at least as
54 * long as the refcount of the inode is positive. As a side effect, the
55 * size of the dcache depends on the inode cache and thus is dynamic.
56 *
57 * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink
58 * resolution to correspond with current state of the code.
59 *
60 * Note that the symlink resolution is not *completely* iterative.
61 * There is still a significant amount of tail- and mid- recursion in
62 * the algorithm. Also, note that <fs>_readlink() is not used in
63 * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink()
64 * may return different results than <fs>_follow_link(). Many virtual
65 * filesystems (including /proc) exhibit this behavior.
66 */
67
68/* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation:
69 * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL
70 * and the name already exists in form of a symlink, try to create the new
71 * name indicated by the symlink. The old code always complained that the
72 * name already exists, due to not following the symlink even if its target
73 * is nonexistent. The new semantics affects also mknod() and link() when
74 * the name is a symlink pointing to a non-existant name.
75 *
76 * I don't know which semantics is the right one, since I have no access
77 * to standards. But I found by trial that HP-UX 9.0 has the full "new"
78 * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the
79 * "old" one. Personally, I think the new semantics is much more logical.
80 * Note that "ln old new" where "new" is a symlink pointing to a non-existing
81 * file does succeed in both HP-UX and SunOs, but not in Solaris
82 * and in the old Linux semantics.
83 */
84
85/* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink
86 * semantics. See the comments in "open_namei" and "do_link" below.
87 *
88 * [10-Sep-98 Alan Modra] Another symlink change.
89 */
90
91/* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks:
92 * inside the path - always follow.
93 * in the last component in creation/removal/renaming - never follow.
94 * if LOOKUP_FOLLOW passed - follow.
95 * if the pathname has trailing slashes - follow.
96 * otherwise - don't follow.
97 * (applied in that order).
98 *
99 * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT
100 * restored for 2.4. This is the last surviving part of old 4.2BSD bug.
101 * During the 2.4 we need to fix the userland stuff depending on it -
102 * hopefully we will be able to get rid of that wart in 2.5. So far only
103 * XEmacs seems to be relying on it...
104 */
105/*
106 * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland)
a11f3a05 107 * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives
1da177e4
LT		 108 * any extra contention...
109 */
110
111/* In order to reduce some races, while at the same time doing additional
112 * checking and hopefully speeding things up, we copy filenames to the
113 * kernel data space before using them..
114 *
115 * POSIX.1 2.4: an empty pathname is invalid (ENOENT).
116 * PATH_MAX includes the nul terminator --RR.
117 */
858119e1 118static int do_getname(const char __user *filename, char *page)
1da177e4
LT		 119{
120 int retval;
121 unsigned long len = PATH_MAX;
122
123 if (!segment_eq(get_fs(), KERNEL_DS)) {
124 if ((unsigned long) filename >= TASK_SIZE)
125 return -EFAULT;
126 if (TASK_SIZE - (unsigned long) filename < PATH_MAX)
127 len = TASK_SIZE - (unsigned long) filename;
128 }
129
130 retval = strncpy_from_user(page, filename, len);
131 if (retval > 0) {
132 if (retval < len)
133 return 0;
134 return -ENAMETOOLONG;
135 } else if (!retval)
136 retval = -ENOENT;
137 return retval;
138}
139
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{
162 if (unlikely(current->audit_context))
163 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;
231
232 if (mask & MAY_WRITE) {
233 umode_t mode = inode->i_mode;
234
235 /*
236 * Nobody gets write access to a read-only fs.
237 */
238 if (IS_RDONLY(inode) &&
239 (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
240 return -EROFS;
241
242 /*
243 * Nobody gets write access to an immutable file.
244 */
245 if (IS_IMMUTABLE(inode))
246 return -EACCES;
247 }
248
249
250 /* Ordinary permission routines do not understand MAY_APPEND. */
251 submask = mask & ~MAY_APPEND;
252 if (inode->i_op && inode->i_op->permission)
253 retval = inode->i_op->permission(inode, submask, nd);
254 else
255 retval = generic_permission(inode, submask, NULL);
256 if (retval)
257 return retval;
258
259 return security_inode_permission(inode, mask, nd);
260}
261
e4543edd
CH		 262/**
263 * vfs_permission - check for access rights to a given path
264 * @nd: lookup result that describes the path
265 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
266 *
267 * Used to check for read/write/execute permissions on a path.
268 * We use "fsuid" for this, letting us set arbitrary permissions
269 * for filesystem access without changing the "normal" uids which
270 * are used for other things.
271 */
272int vfs_permission(struct nameidata *nd, int mask)
273{
274 return permission(nd->dentry->d_inode, mask, nd);
275}
276
8c744fb8
CH		 277/**
278 * file_permission - check for additional access rights to a given file
279 * @file: file to check access rights for
280 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
281 *
282 * Used to check for read/write/execute permissions on an already opened
283 * file.
284 *
285 * Note:
286 * Do not use this function in new code. All access checks should
287 * be done using vfs_permission().
288 */
289int file_permission(struct file *file, int mask)
290{
291 return permission(file->f_dentry->d_inode, mask, NULL);
292}
293
1da177e4
LT		 294/*
295 * get_write_access() gets write permission for a file.
296 * put_write_access() releases this write permission.
297 * This is used for regular files.
298 * We cannot support write (and maybe mmap read-write shared) accesses and
299 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
300 * can have the following values:
301 * 0: no writers, no VM_DENYWRITE mappings
302 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
303 * > 0: (i_writecount) users are writing to the file.
304 *
305 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
306 * except for the cases where we don't hold i_writecount yet. Then we need to
307 * use {get,deny}_write_access() - these functions check the sign and refuse
308 * to do the change if sign is wrong. Exclusion between them is provided by
309 * the inode->i_lock spinlock.
310 */
311
312int get_write_access(struct inode * inode)
313{
314 spin_lock(&inode->i_lock);
315 if (atomic_read(&inode->i_writecount) < 0) {
316 spin_unlock(&inode->i_lock);
317 return -ETXTBSY;
318 }
319 atomic_inc(&inode->i_writecount);
320 spin_unlock(&inode->i_lock);
321
322 return 0;
323}
324
325int deny_write_access(struct file * file)
326{
327 struct inode *inode = file->f_dentry->d_inode;
328
329 spin_lock(&inode->i_lock);
330 if (atomic_read(&inode->i_writecount) > 0) {
331 spin_unlock(&inode->i_lock);
332 return -ETXTBSY;
333 }
334 atomic_dec(&inode->i_writecount);
335 spin_unlock(&inode->i_lock);
336
337 return 0;
338}
339
340void path_release(struct nameidata *nd)
341{
342 dput(nd->dentry);
343 mntput(nd->mnt);
344}
345
346/*
347 * umount() mustn't call path_release()/mntput() as that would clear
348 * mnt_expiry_mark
349 */
350void path_release_on_umount(struct nameidata *nd)
351{
352 dput(nd->dentry);
751c404b 353 mntput_no_expire(nd->mnt);
1da177e4
LT		 354}
355
834f2a4a
TM		 356/**
357 * release_open_intent - free up open intent resources
358 * @nd: pointer to nameidata
359 */
360void release_open_intent(struct nameidata *nd)
361{
362 if (nd->intent.open.file->f_dentry == NULL)
363 put_filp(nd->intent.open.file);
364 else
365 fput(nd->intent.open.file);
366}
367
1da177e4
LT		 368/*
369 * Internal lookup() using the new generic dcache.
370 * SMP-safe
371 */
372static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd)
373{
374 struct dentry * dentry = __d_lookup(parent, name);
375
376 /* lockess __d_lookup may fail due to concurrent d_move()
377 * in some unrelated directory, so try with d_lookup
378 */
379 if (!dentry)
380 dentry = d_lookup(parent, name);
381
382 if (dentry && dentry->d_op && dentry->d_op->d_revalidate) {
383 if (!dentry->d_op->d_revalidate(dentry, nd) && !d_invalidate(dentry)) {
384 dput(dentry);
385 dentry = NULL;
386 }
387 }
388 return dentry;
389}
390
391/*
392 * Short-cut version of permission(), for calling by
393 * path_walk(), when dcache lock is held. Combines parts
394 * of permission() and generic_permission(), and tests ONLY for
395 * MAY_EXEC permission.
396 *
397 * If appropriate, check DAC only. If not appropriate, or
398 * short-cut DAC fails, then call permission() to do more
399 * complete permission check.
400 */
858119e1 401static int exec_permission_lite(struct inode *inode,
1da177e4
LT		 402 struct nameidata *nd)
403{
404 umode_t mode = inode->i_mode;
405
406 if (inode->i_op && inode->i_op->permission)
407 return -EAGAIN;
408
409 if (current->fsuid == inode->i_uid)
410 mode >>= 6;
411 else if (in_group_p(inode->i_gid))
412 mode >>= 3;
413
414 if (mode & MAY_EXEC)
415 goto ok;
416
417 if ((inode->i_mode & S_IXUGO) && capable(CAP_DAC_OVERRIDE))
418 goto ok;
419
420 if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_OVERRIDE))
421 goto ok;
422
423 if (S_ISDIR(inode->i_mode) && capable(CAP_DAC_READ_SEARCH))
424 goto ok;
425
426 return -EACCES;
427ok:
428 return security_inode_permission(inode, MAY_EXEC, nd);
429}
430
431/*
432 * This is called when everything else fails, and we actually have
433 * to go to the low-level filesystem to find out what we should do..
434 *
435 * We get the directory semaphore, and after getting that we also
436 * make sure that nobody added the entry to the dcache in the meantime..
437 * SMP-safe
438 */
439static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, struct nameidata *nd)
440{
441 struct dentry * result;
442 struct inode *dir = parent->d_inode;
443
1b1dcc1b 444 mutex_lock(&dir->i_mutex);
1da177e4
LT		 445 /*
446 * First re-do the cached lookup just in case it was created
447 * while we waited for the directory semaphore..
448 *
449 * FIXME! This could use version numbering or similar to
450 * avoid unnecessary cache lookups.
451 *
452 * The "dcache_lock" is purely to protect the RCU list walker
453 * from concurrent renames at this point (we mustn't get false
454 * negatives from the RCU list walk here, unlike the optimistic
455 * fast walk).
456 *
457 * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup
458 */
459 result = d_lookup(parent, name);
460 if (!result) {
461 struct dentry * dentry = d_alloc(parent, name);
462 result = ERR_PTR(-ENOMEM);
463 if (dentry) {
464 result = dir->i_op->lookup(dir, dentry, nd);
465 if (result)
466 dput(dentry);
467 else
468 result = dentry;
469 }
1b1dcc1b 470 mutex_unlock(&dir->i_mutex);
1da177e4
LT		 471 return result;
472 }
473
474 /*
475 * Uhhuh! Nasty case: the cache was re-populated while
476 * we waited on the semaphore. Need to revalidate.
477 */
1b1dcc1b 478 mutex_unlock(&dir->i_mutex);
1da177e4
LT		 479 if (result->d_op && result->d_op->d_revalidate) {
480 if (!result->d_op->d_revalidate(result, nd) && !d_invalidate(result)) {
481 dput(result);
482 result = ERR_PTR(-ENOENT);
483 }
484 }
485 return result;
486}
487
488static int __emul_lookup_dentry(const char *, struct nameidata *);
489
490/* SMP-safe */
f1662356 491static __always_inline int
1da177e4
LT		 492walk_init_root(const char *name, struct nameidata *nd)
493{
494 read_lock(&current->fs->lock);
495 if (current->fs->altroot && !(nd->flags & LOOKUP_NOALT)) {
496 nd->mnt = mntget(current->fs->altrootmnt);
497 nd->dentry = dget(current->fs->altroot);
498 read_unlock(&current->fs->lock);
499 if (__emul_lookup_dentry(name,nd))
500 return 0;
501 read_lock(&current->fs->lock);
502 }
503 nd->mnt = mntget(current->fs->rootmnt);
504 nd->dentry = dget(current->fs->root);
505 read_unlock(&current->fs->lock);
506 return 1;
507}
508
f1662356 509static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link)
1da177e4
LT		 510{
511 int res = 0;
512 char *name;
513 if (IS_ERR(link))
514 goto fail;
515
516 if (*link == '/') {
517 path_release(nd);
518 if (!walk_init_root(link, nd))
519 /* weird __emul_prefix() stuff did it */
520 goto out;
521 }
522 res = link_path_walk(link, nd);
523out:
524 if (nd->depth || res || nd->last_type!=LAST_NORM)
525 return res;
526 /*
527 * If it is an iterative symlinks resolution in open_namei() we
528 * have to copy the last component. And all that crap because of
529 * bloody create() on broken symlinks. Furrfu...
530 */
531 name = __getname();
532 if (unlikely(!name)) {
533 path_release(nd);
534 return -ENOMEM;
535 }
536 strcpy(name, nd->last.name);
537 nd->last.name = name;
538 return 0;
539fail:
540 path_release(nd);
541 return PTR_ERR(link);
542}
543
90ebe565
AV		 544struct path {
545 struct vfsmount *mnt;
546 struct dentry *dentry;
547};
548
051d3812
IK		 549static inline void dput_path(struct path *path, struct nameidata *nd)
550{
551 dput(path->dentry);
552 if (path->mnt != nd->mnt)
553 mntput(path->mnt);
554}
555
556static inline void path_to_nameidata(struct path *path, struct nameidata *nd)
557{
558 dput(nd->dentry);
559 if (nd->mnt != path->mnt)
560 mntput(nd->mnt);
561 nd->mnt = path->mnt;
562 nd->dentry = path->dentry;
563}
564
f1662356 565static __always_inline int __do_follow_link(struct path *path, struct nameidata *nd)
1da177e4
LT		 566{
567 int error;
cc314eef 568 void *cookie;
cd4e91d3 569 struct dentry *dentry = path->dentry;
1da177e4 570
d671a1cb 571 touch_atime(path->mnt, dentry);
1da177e4 572 nd_set_link(nd, NULL);
cd4e91d3 573
051d3812
IK		 574 if (path->mnt != nd->mnt) {
575 path_to_nameidata(path, nd);
576 dget(dentry);
577 }
578 mntget(path->mnt);
cc314eef
LT		 579 cookie = dentry->d_inode->i_op->follow_link(dentry, nd);
580 error = PTR_ERR(cookie);
581 if (!IS_ERR(cookie)) {
1da177e4 582 char *s = nd_get_link(nd);
cc314eef 583 error = 0;
1da177e4
LT		 584 if (s)
585 error = __vfs_follow_link(nd, s);
586 if (dentry->d_inode->i_op->put_link)
cc314eef 587 dentry->d_inode->i_op->put_link(dentry, nd, cookie);
1da177e4 588 }
cd4e91d3
AV		 589 dput(dentry);
590 mntput(path->mnt);
1da177e4
LT		 591
592 return error;
593}
594
595/*
596 * This limits recursive symlink follows to 8, while
597 * limiting consecutive symlinks to 40.
598 *
599 * Without that kind of total limit, nasty chains of consecutive
600 * symlinks can cause almost arbitrarily long lookups.
601 */
90ebe565 602static inline int do_follow_link(struct path *path, struct nameidata *nd)
1da177e4
LT		 603{
604 int err = -ELOOP;
605 if (current->link_count >= MAX_NESTED_LINKS)
606 goto loop;
607 if (current->total_link_count >= 40)
608 goto loop;
609 BUG_ON(nd->depth >= MAX_NESTED_LINKS);
610 cond_resched();
90ebe565 611 err = security_inode_follow_link(path->dentry, nd);
1da177e4
LT		 612 if (err)
613 goto loop;
614 current->link_count++;
615 current->total_link_count++;
616 nd->depth++;
cd4e91d3 617 err = __do_follow_link(path, nd);
839d9f93
AV		 618 current->link_count--;
619 nd->depth--;
1da177e4
LT		 620 return err;
621loop:
09dd17d3 622 dput_path(path, nd);
839d9f93 623 path_release(nd);
1da177e4
LT		 624 return err;
625}
626
627int follow_up(struct vfsmount **mnt, struct dentry **dentry)
628{
629 struct vfsmount *parent;
630 struct dentry *mountpoint;
631 spin_lock(&vfsmount_lock);
632 parent=(*mnt)->mnt_parent;
633 if (parent == *mnt) {
634 spin_unlock(&vfsmount_lock);
635 return 0;
636 }
637 mntget(parent);
638 mountpoint=dget((*mnt)->mnt_mountpoint);
639 spin_unlock(&vfsmount_lock);
640 dput(*dentry);
641 *dentry = mountpoint;
642 mntput(*mnt);
643 *mnt = parent;
644 return 1;
645}
646
647/* no need for dcache_lock, as serialization is taken care in
648 * namespace.c
649 */
463ffb2e
AV		 650static int __follow_mount(struct path *path)
651{
652 int res = 0;
653 while (d_mountpoint(path->dentry)) {
654 struct vfsmount *mounted = lookup_mnt(path->mnt, path->dentry);
655 if (!mounted)
656 break;
657 dput(path->dentry);
658 if (res)
659 mntput(path->mnt);
660 path->mnt = mounted;
661 path->dentry = dget(mounted->mnt_root);
662 res = 1;
663 }
664 return res;
665}
666
58c465eb 667static void follow_mount(struct vfsmount **mnt, struct dentry **dentry)
1da177e4 668{
1da177e4
LT		 669 while (d_mountpoint(*dentry)) {
670 struct vfsmount *mounted = lookup_mnt(*mnt, *dentry);
671 if (!mounted)
672 break;
58c465eb 673 dput(*dentry);
1da177e4
LT		 674 mntput(*mnt);
675 *mnt = mounted;
1da177e4 676 *dentry = dget(mounted->mnt_root);
1da177e4 677 }
1da177e4
LT		 678}
679
680/* no need for dcache_lock, as serialization is taken care in
681 * namespace.c
682 */
e13b210f 683int follow_down(struct vfsmount **mnt, struct dentry **dentry)
1da177e4
LT		 684{
685 struct vfsmount *mounted;
686
687 mounted = lookup_mnt(*mnt, *dentry);
688 if (mounted) {
e13b210f 689 dput(*dentry);
1da177e4
LT		 690 mntput(*mnt);
691 *mnt = mounted;
1da177e4
LT		 692 *dentry = dget(mounted->mnt_root);
693 return 1;
694 }
695 return 0;
696}
697
f1662356 698static __always_inline void follow_dotdot(struct nameidata *nd)
1da177e4
LT		 699{
700 while(1) {
701 struct vfsmount *parent;
58c465eb 702 struct dentry *old = nd->dentry;
1da177e4
LT		 703
704 read_lock(&current->fs->lock);
58c465eb
AV		 705 if (nd->dentry == current->fs->root &&
706 nd->mnt == current->fs->rootmnt) {
1da177e4
LT		 707 read_unlock(&current->fs->lock);
708 break;
709 }
710 read_unlock(&current->fs->lock);
711 spin_lock(&dcache_lock);
58c465eb
AV		 712 if (nd->dentry != nd->mnt->mnt_root) {
713 nd->dentry = dget(nd->dentry->d_parent);
1da177e4
LT		 714 spin_unlock(&dcache_lock);
715 dput(old);
716 break;
717 }
718 spin_unlock(&dcache_lock);
719 spin_lock(&vfsmount_lock);
58c465eb
AV		 720 parent = nd->mnt->mnt_parent;
721 if (parent == nd->mnt) {
1da177e4
LT		 722 spin_unlock(&vfsmount_lock);
723 break;
724 }
725 mntget(parent);
58c465eb 726 nd->dentry = dget(nd->mnt->mnt_mountpoint);
1da177e4
LT		 727 spin_unlock(&vfsmount_lock);
728 dput(old);
58c465eb
AV		 729 mntput(nd->mnt);
730 nd->mnt = parent;
1da177e4 731 }
58c465eb 732 follow_mount(&nd->mnt, &nd->dentry);
1da177e4
LT		 733}
734
1da177e4
LT		 735/*
736 * It's more convoluted than I'd like it to be, but... it's still fairly
737 * small and for now I'd prefer to have fast path as straight as possible.
738 * It _is_ time-critical.
739 */
740static int do_lookup(struct nameidata *nd, struct qstr *name,
741 struct path *path)
742{
743 struct vfsmount *mnt = nd->mnt;
744 struct dentry *dentry = __d_lookup(nd->dentry, name);
745
746 if (!dentry)
747 goto need_lookup;
748 if (dentry->d_op && dentry->d_op->d_revalidate)
749 goto need_revalidate;
750done:
751 path->mnt = mnt;
752 path->dentry = dentry;
634ee701 753 __follow_mount(path);
1da177e4
LT		 754 return 0;
755
756need_lookup:
757 dentry = real_lookup(nd->dentry, name, nd);
758 if (IS_ERR(dentry))
759 goto fail;
760 goto done;
761
762need_revalidate:
763 if (dentry->d_op->d_revalidate(dentry, nd))
764 goto done;
765 if (d_invalidate(dentry))
766 goto done;
767 dput(dentry);
768 goto need_lookup;
769
770fail:
771 return PTR_ERR(dentry);
772}
773
774/*
775 * Name resolution.
ea3834d9
PM		 776 * This is the basic name resolution function, turning a pathname into
777 * the final dentry. We expect 'base' to be positive and a directory.
1da177e4 778 *
ea3834d9
PM		 779 * Returns 0 and nd will have valid dentry and mnt on success.
780 * Returns error and drops reference to input namei data on failure.
1da177e4
LT		 781 */
782static fastcall int __link_path_walk(const char * name, struct nameidata *nd)
783{
784 struct path next;
785 struct inode *inode;
786 int err;
787 unsigned int lookup_flags = nd->flags;
788
789 while (*name=='/')
790 name++;
791 if (!*name)
792 goto return_reval;
793
794 inode = nd->dentry->d_inode;
795 if (nd->depth)
f55eab82 796 lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE);
1da177e4
LT		 797
798 /* At this point we know we have a real path component. */
799 for(;;) {
800 unsigned long hash;
801 struct qstr this;
802 unsigned int c;
803
cdce5d6b 804 nd->flags |= LOOKUP_CONTINUE;
1da177e4 805 err = exec_permission_lite(inode, nd);
e4543edd
CH		 806 if (err == -EAGAIN)
807 err = vfs_permission(nd, MAY_EXEC);
1da177e4
LT		 808 if (err)
809 break;
810
811 this.name = name;
812 c = *(const unsigned char *)name;
813
814 hash = init_name_hash();
815 do {
816 name++;
817 hash = partial_name_hash(c, hash);
818 c = *(const unsigned char *)name;
819 } while (c && (c != '/'));
820 this.len = name - (const char *) this.name;
821 this.hash = end_name_hash(hash);
822
823 /* remove trailing slashes? */
824 if (!c)
825 goto last_component;
826 while (*++name == '/');
827 if (!*name)
828 goto last_with_slashes;
829
830 /*
831 * "." and ".." are special - ".." especially so because it has
832 * to be able to know about the current root directory and
833 * parent relationships.
834 */
835 if (this.name[0] == '.') switch (this.len) {
836 default:
837 break;
838 case 2:
839 if (this.name[1] != '.')
840 break;
58c465eb 841 follow_dotdot(nd);
1da177e4
LT		 842 inode = nd->dentry->d_inode;
843 /* fallthrough */
844 case 1:
845 continue;
846 }
847 /*
848 * See if the low-level filesystem might want
849 * to use its own hash..
850 */
851 if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
852 err = nd->dentry->d_op->d_hash(nd->dentry, &this);
853 if (err < 0)
854 break;
855 }
1da177e4
LT		 856 /* This does the actual lookups.. */
857 err = do_lookup(nd, &this, &next);
858 if (err)
859 break;
1da177e4
LT		 860
861 err = -ENOENT;
862 inode = next.dentry->d_inode;
863 if (!inode)
864 goto out_dput;
865 err = -ENOTDIR;
866 if (!inode->i_op)
867 goto out_dput;
868
869 if (inode->i_op->follow_link) {
90ebe565 870 err = do_follow_link(&next, nd);
1da177e4
LT		 871 if (err)
872 goto return_err;
873 err = -ENOENT;
874 inode = nd->dentry->d_inode;
875 if (!inode)
876 break;
877 err = -ENOTDIR;
878 if (!inode->i_op)
879 break;
09dd17d3
MS		 880 } else
881 path_to_nameidata(&next, nd);
1da177e4
LT		 882 err = -ENOTDIR;
883 if (!inode->i_op->lookup)
884 break;
885 continue;
886 /* here ends the main loop */
887
888last_with_slashes:
889 lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
890last_component:
f55eab82
TM		 891 /* Clear LOOKUP_CONTINUE iff it was previously unset */
892 nd->flags &= lookup_flags | ~LOOKUP_CONTINUE;
1da177e4
LT		 893 if (lookup_flags & LOOKUP_PARENT)
894 goto lookup_parent;
895 if (this.name[0] == '.') switch (this.len) {
896 default:
897 break;
898 case 2:
899 if (this.name[1] != '.')
900 break;
58c465eb 901 follow_dotdot(nd);
1da177e4
LT		 902 inode = nd->dentry->d_inode;
903 /* fallthrough */
904 case 1:
905 goto return_reval;
906 }
907 if (nd->dentry->d_op && nd->dentry->d_op->d_hash) {
908 err = nd->dentry->d_op->d_hash(nd->dentry, &this);
909 if (err < 0)
910 break;
911 }
912 err = do_lookup(nd, &this, &next);
913 if (err)
914 break;
1da177e4
LT		 915 inode = next.dentry->d_inode;
916 if ((lookup_flags & LOOKUP_FOLLOW)
917 && inode && inode->i_op && inode->i_op->follow_link) {
90ebe565 918 err = do_follow_link(&next, nd);
1da177e4
LT		 919 if (err)
920 goto return_err;
921 inode = nd->dentry->d_inode;
09dd17d3
MS		 922 } else
923 path_to_nameidata(&next, nd);
1da177e4
LT		 924 err = -ENOENT;
925 if (!inode)
926 break;
927 if (lookup_flags & LOOKUP_DIRECTORY) {
928 err = -ENOTDIR;
929 if (!inode->i_op || !inode->i_op->lookup)
930 break;
931 }
932 goto return_base;
933lookup_parent:
934 nd->last = this;
935 nd->last_type = LAST_NORM;
936 if (this.name[0] != '.')
937 goto return_base;
938 if (this.len == 1)
939 nd->last_type = LAST_DOT;
940 else if (this.len == 2 && this.name[1] == '.')
941 nd->last_type = LAST_DOTDOT;
942 else
943 goto return_base;
944return_reval:
945 /*
946 * We bypassed the ordinary revalidation routines.
947 * We may need to check the cached dentry for staleness.
948 */
949 if (nd->dentry && nd->dentry->d_sb &&
950 (nd->dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) {
951 err = -ESTALE;
952 /* Note: we do not d_invalidate() */
953 if (!nd->dentry->d_op->d_revalidate(nd->dentry, nd))
954 break;
955 }
956return_base:
957 return 0;
958out_dput:
09dd17d3 959 dput_path(&next, nd);
1da177e4
LT		 960 break;
961 }
962 path_release(nd);
963return_err:
964 return err;
965}
966
967/*
968 * Wrapper to retry pathname resolution whenever the underlying
969 * file system returns an ESTALE.
970 *
971 * Retry the whole path once, forcing real lookup requests
972 * instead of relying on the dcache.
973 */
974int fastcall link_path_walk(const char *name, struct nameidata *nd)
975{
976 struct nameidata save = *nd;
977 int result;
978
979 /* make sure the stuff we saved doesn't go away */
980 dget(save.dentry);
981 mntget(save.mnt);
982
983 result = __link_path_walk(name, nd);
984 if (result == -ESTALE) {
985 *nd = save;
986 dget(nd->dentry);
987 mntget(nd->mnt);
988 nd->flags |= LOOKUP_REVAL;
989 result = __link_path_walk(name, nd);
990 }
991
992 dput(save.dentry);
993 mntput(save.mnt);
994
995 return result;
996}
997
998int fastcall path_walk(const char * name, struct nameidata *nd)
999{
1000 current->total_link_count = 0;
1001 return link_path_walk(name, nd);
1002}
1003
ea3834d9
PM		 1004/*
1005 * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if
1006 * everything is done. Returns 0 and drops input nd, if lookup failed;
1007 */
1da177e4
LT		 1008static int __emul_lookup_dentry(const char *name, struct nameidata *nd)
1009{
1010 if (path_walk(name, nd))
1011 return 0; /* something went wrong... */
1012
1013 if (!nd->dentry->d_inode || S_ISDIR(nd->dentry->d_inode->i_mode)) {
1014 struct dentry *old_dentry = nd->dentry;
1015 struct vfsmount *old_mnt = nd->mnt;
1016 struct qstr last = nd->last;
1017 int last_type = nd->last_type;
1018 /*
1019 * NAME was not found in alternate root or it's a directory. Try to find
1020 * it in the normal root:
1021 */
1022 nd->last_type = LAST_ROOT;
1023 read_lock(&current->fs->lock);
1024 nd->mnt = mntget(current->fs->rootmnt);
1025 nd->dentry = dget(current->fs->root);
1026 read_unlock(&current->fs->lock);
1027 if (path_walk(name, nd) == 0) {
1028 if (nd->dentry->d_inode) {
1029 dput(old_dentry);
1030 mntput(old_mnt);
1031 return 1;
1032 }
1033 path_release(nd);
1034 }
1035 nd->dentry = old_dentry;
1036 nd->mnt = old_mnt;
1037 nd->last = last;
1038 nd->last_type = last_type;
1039 }
1040 return 1;
1041}
1042
1043void set_fs_altroot(void)
1044{
1045 char *emul = __emul_prefix();
1046 struct nameidata nd;
1047 struct vfsmount *mnt = NULL, *oldmnt;
1048 struct dentry *dentry = NULL, *olddentry;
1049 int err;
1050
1051 if (!emul)
1052 goto set_it;
1053 err = path_lookup(emul, LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_NOALT, &nd);
1054 if (!err) {
1055 mnt = nd.mnt;
1056 dentry = nd.dentry;
1057 }
1058set_it:
1059 write_lock(&current->fs->lock);
1060 oldmnt = current->fs->altrootmnt;
1061 olddentry = current->fs->altroot;
1062 current->fs->altrootmnt = mnt;
1063 current->fs->altroot = dentry;
1064 write_unlock(&current->fs->lock);
1065 if (olddentry) {
1066 dput(olddentry);
1067 mntput(oldmnt);
1068 }
1069}
1070
ea3834d9 1071/* Returns 0 and nd will be valid on success; Retuns error, otherwise. */
5590ff0d
UD		 1072static int fastcall do_path_lookup(int dfd, const char *name,
1073 unsigned int flags, struct nameidata *nd)
1da177e4 1074{
ea3834d9 1075 int retval = 0;
170aa3d0
UD		 1076 int fput_needed;
1077 struct file *file;
1da177e4
LT		 1078
1079 nd->last_type = LAST_ROOT; /* if there are only slashes... */
1080 nd->flags = flags;
1081 nd->depth = 0;
1082
1da177e4 1083 if (*name=='/') {
6d09bb62 1084 read_lock(&current->fs->lock);
1da177e4
LT		 1085 if (current->fs->altroot && !(nd->flags & LOOKUP_NOALT)) {
1086 nd->mnt = mntget(current->fs->altrootmnt);
1087 nd->dentry = dget(current->fs->altroot);
1088 read_unlock(&current->fs->lock);
1089 if (__emul_lookup_dentry(name,nd))
ea3834d9 1090 goto out; /* found in altroot */
1da177e4
LT		 1091 read_lock(&current->fs->lock);
1092 }
1093 nd->mnt = mntget(current->fs->rootmnt);
1094 nd->dentry = dget(current->fs->root);
6d09bb62 1095 read_unlock(&current->fs->lock);
5590ff0d 1096 } else if (dfd == AT_FDCWD) {
6d09bb62 1097 read_lock(&current->fs->lock);
1da177e4
LT		 1098 nd->mnt = mntget(current->fs->pwdmnt);
1099 nd->dentry = dget(current->fs->pwd);
6d09bb62 1100 read_unlock(&current->fs->lock);
5590ff0d 1101 } else {
5590ff0d
UD		 1102 struct dentry *dentry;
1103
1104 file = fget_light(dfd, &fput_needed);
170aa3d0
UD		 1105 retval = -EBADF;
1106 if (!file)
6d09bb62 1107 goto out_fail;
5590ff0d
UD		 1108
1109 dentry = file->f_dentry;
1110
170aa3d0
UD		 1111 retval = -ENOTDIR;
1112 if (!S_ISDIR(dentry->d_inode->i_mode))
6d09bb62 1113 goto fput_fail;
5590ff0d
UD		 1114
1115 retval = file_permission(file, MAY_EXEC);
170aa3d0 1116 if (retval)
6d09bb62 1117 goto fput_fail;
5590ff0d
UD		 1118
1119 nd->mnt = mntget(file->f_vfsmnt);
1120 nd->dentry = dget(dentry);
1121
1122 fput_light(file, fput_needed);
1da177e4 1123 }
1da177e4
LT		 1124 current->total_link_count = 0;
1125 retval = link_path_walk(name, nd);
ea3834d9 1126out:
3bc8414b
S		 1127 if (likely(retval == 0)) {
1128 if (unlikely(current->audit_context && nd && nd->dentry &&
1129 nd->dentry->d_inode))
9c937dcc 1130 audit_inode(name, nd->dentry->d_inode);
3bc8414b 1131 }
6d09bb62 1132out_fail:
170aa3d0
UD		 1133 return retval;
1134
6d09bb62 1135fput_fail:
170aa3d0 1136 fput_light(file, fput_needed);
6d09bb62 1137 goto out_fail;
1da177e4
LT		 1138}
1139
5590ff0d
UD		 1140int fastcall path_lookup(const char *name, unsigned int flags,
1141 struct nameidata *nd)
1142{
1143 return do_path_lookup(AT_FDCWD, name, flags, nd);
1144}
1145
1146static int __path_lookup_intent_open(int dfd, const char *name,
1147 unsigned int lookup_flags, struct nameidata *nd,
1148 int open_flags, int create_mode)
834f2a4a
TM		 1149{
1150 struct file *filp = get_empty_filp();
1151 int err;
1152
1153 if (filp == NULL)
1154 return -ENFILE;
1155 nd->intent.open.file = filp;
1156 nd->intent.open.flags = open_flags;
1157 nd->intent.open.create_mode = create_mode;
5590ff0d 1158 err = do_path_lookup(dfd, name, lookup_flags|LOOKUP_OPEN, nd);
834f2a4a
TM		 1159 if (IS_ERR(nd->intent.open.file)) {
1160 if (err == 0) {
1161 err = PTR_ERR(nd->intent.open.file);
1162 path_release(nd);
1163 }
1164 } else if (err != 0)
1165 release_open_intent(nd);
1166 return err;
1167}
1168
1169/**
1170 * path_lookup_open - lookup a file path with open intent
7045f37b 1171 * @dfd: the directory to use as base, or AT_FDCWD
834f2a4a
TM		 1172 * @name: pointer to file name
1173 * @lookup_flags: lookup intent flags
1174 * @nd: pointer to nameidata
1175 * @open_flags: open intent flags
1176 */
5590ff0d 1177int path_lookup_open(int dfd, const char *name, unsigned int lookup_flags,
834f2a4a
TM		 1178 struct nameidata *nd, int open_flags)
1179{
5590ff0d 1180 return __path_lookup_intent_open(dfd, name, lookup_flags, nd,
834f2a4a
TM		 1181 open_flags, 0);
1182}
1183
1184/**
1185 * path_lookup_create - lookup a file path with open + create intent
7045f37b 1186 * @dfd: the directory to use as base, or AT_FDCWD
834f2a4a
TM		 1187 * @name: pointer to file name
1188 * @lookup_flags: lookup intent flags
1189 * @nd: pointer to nameidata
1190 * @open_flags: open intent flags
1191 * @create_mode: create intent flags
1192 */
5590ff0d
UD		 1193static int path_lookup_create(int dfd, const char *name,
1194 unsigned int lookup_flags, struct nameidata *nd,
1195 int open_flags, int create_mode)
834f2a4a 1196{
5590ff0d
UD		 1197 return __path_lookup_intent_open(dfd, name, lookup_flags|LOOKUP_CREATE,
1198 nd, open_flags, create_mode);
834f2a4a
TM		 1199}
1200
1201int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags,
1202 struct nameidata *nd, int open_flags)
1203{
1204 char *tmp = getname(name);
1205 int err = PTR_ERR(tmp);
1206
1207 if (!IS_ERR(tmp)) {
5590ff0d 1208 err = __path_lookup_intent_open(AT_FDCWD, tmp, lookup_flags, nd, open_flags, 0);
834f2a4a
TM		 1209 putname(tmp);
1210 }
1211 return err;
1212}
1213
1da177e4
LT		 1214/*
1215 * Restricted form of lookup. Doesn't follow links, single-component only,
1216 * needs parent already locked. Doesn't follow mounts.
1217 * SMP-safe.
1218 */
1219static struct dentry * __lookup_hash(struct qstr *name, struct dentry * base, struct nameidata *nd)
1220{
1221 struct dentry * dentry;
1222 struct inode *inode;
1223 int err;
1224
1225 inode = base->d_inode;
1226 err = permission(inode, MAY_EXEC, nd);
1227 dentry = ERR_PTR(err);
1228 if (err)
1229 goto out;
1230
1231 /*
1232 * See if the low-level filesystem might want
1233 * to use its own hash..
1234 */
1235 if (base->d_op && base->d_op->d_hash) {
1236 err = base->d_op->d_hash(base, name);
1237 dentry = ERR_PTR(err);
1238 if (err < 0)
1239 goto out;
1240 }
1241
1242 dentry = cached_lookup(base, name, nd);
1243 if (!dentry) {
1244 struct dentry *new = d_alloc(base, name);
1245 dentry = ERR_PTR(-ENOMEM);
1246 if (!new)
1247 goto out;
1248 dentry = inode->i_op->lookup(inode, new, nd);
1249 if (!dentry)
1250 dentry = new;
1251 else
1252 dput(new);
1253 }
1254out:
1255 return dentry;
1256}
1257
a244e169 1258static struct dentry *lookup_hash(struct nameidata *nd)
1da177e4 1259{
49705b77 1260 return __lookup_hash(&nd->last, nd->dentry, nd);
1da177e4
LT		 1261}
1262
1263/* SMP-safe */
1264struct dentry * lookup_one_len(const char * name, struct dentry * base, int len)
1265{
1266 unsigned long hash;
1267 struct qstr this;
1268 unsigned int c;
1269
1270 this.name = name;
1271 this.len = len;
1272 if (!len)
1273 goto access;
1274
1275 hash = init_name_hash();
1276 while (len--) {
1277 c = *(const unsigned char *)name++;
1278 if (c == '/' || c == '\0')
1279 goto access;
1280 hash = partial_name_hash(c, hash);
1281 }
1282 this.hash = end_name_hash(hash);
1283
49705b77 1284 return __lookup_hash(&this, base, NULL);
1da177e4
LT		 1285access:
1286 return ERR_PTR(-EACCES);
1287}
1288
1289/*
1290 * namei()
1291 *
1292 * is used by most simple commands to get the inode of a specified name.
1293 * Open, link etc use their own routines, but this is enough for things
1294 * like 'chmod' etc.
1295 *
1296 * namei exists in two versions: namei/lnamei. The only difference is
1297 * that namei follows links, while lnamei does not.
1298 * SMP-safe
1299 */
5590ff0d
UD		 1300int fastcall __user_walk_fd(int dfd, const char __user *name, unsigned flags,
1301 struct nameidata *nd)
1da177e4
LT		 1302{
1303 char *tmp = getname(name);
1304 int err = PTR_ERR(tmp);
1305
1306 if (!IS_ERR(tmp)) {
5590ff0d 1307 err = do_path_lookup(dfd, tmp, flags, nd);
1da177e4
LT		 1308 putname(tmp);
1309 }
1310 return err;
1311}
1312
5590ff0d
UD		 1313int fastcall __user_walk(const char __user *name, unsigned flags, struct nameidata *nd)
1314{
1315 return __user_walk_fd(AT_FDCWD, name, flags, nd);
1316}
1317
1da177e4
LT		 1318/*
1319 * It's inline, so penalty for filesystems that don't use sticky bit is
1320 * minimal.
1321 */
1322static inline int check_sticky(struct inode *dir, struct inode *inode)
1323{
1324 if (!(dir->i_mode & S_ISVTX))
1325 return 0;
1326 if (inode->i_uid == current->fsuid)
1327 return 0;
1328 if (dir->i_uid == current->fsuid)
1329 return 0;
1330 return !capable(CAP_FOWNER);
1331}
1332
1333/*
1334 * Check whether we can remove a link victim from directory dir, check
1335 * whether the type of victim is right.
1336 * 1. We can't do it if dir is read-only (done in permission())
1337 * 2. We should have write and exec permissions on dir
1338 * 3. We can't remove anything from append-only dir
1339 * 4. We can't do anything with immutable dir (done in permission())
1340 * 5. If the sticky bit on dir is set we should either
1341 * a. be owner of dir, or
1342 * b. be owner of victim, or
1343 * c. have CAP_FOWNER capability
1344 * 6. If the victim is append-only or immutable we can't do antyhing with
1345 * links pointing to it.
1346 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
1347 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
1348 * 9. We can't remove a root or mountpoint.
1349 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
1350 * nfs_async_unlink().
1351 */
858119e1 1352static int may_delete(struct inode *dir,struct dentry *victim,int isdir)
1da177e4
LT		 1353{
1354 int error;
1355
1356 if (!victim->d_inode)
1357 return -ENOENT;
1358
1359 BUG_ON(victim->d_parent->d_inode != dir);
73d3ec5a 1360 audit_inode_child(victim->d_name.name, victim->d_inode, dir);
1da177e4
LT		 1361
1362 error = permission(dir,MAY_WRITE | MAY_EXEC, NULL);
1363 if (error)
1364 return error;
1365 if (IS_APPEND(dir))
1366 return -EPERM;
1367 if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)||
1368 IS_IMMUTABLE(victim->d_inode))
1369 return -EPERM;
1370 if (isdir) {
1371 if (!S_ISDIR(victim->d_inode->i_mode))
1372 return -ENOTDIR;
1373 if (IS_ROOT(victim))
1374 return -EBUSY;
1375 } else if (S_ISDIR(victim->d_inode->i_mode))
1376 return -EISDIR;
1377 if (IS_DEADDIR(dir))
1378 return -ENOENT;
1379 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
1380 return -EBUSY;
1381 return 0;
1382}
1383
1384/* Check whether we can create an object with dentry child in directory
1385 * dir.
1386 * 1. We can't do it if child already exists (open has special treatment for
1387 * this case, but since we are inlined it's OK)
1388 * 2. We can't do it if dir is read-only (done in permission())
1389 * 3. We should have write and exec permissions on dir
1390 * 4. We can't do it if dir is immutable (done in permission())
1391 */
1392static inline int may_create(struct inode *dir, struct dentry *child,
1393 struct nameidata *nd)
1394{
1395 if (child->d_inode)
1396 return -EEXIST;
1397 if (IS_DEADDIR(dir))
1398 return -ENOENT;
1399 return permission(dir,MAY_WRITE | MAY_EXEC, nd);
1400}
1401
1402/*
1da177e4
LT		 1403 * O_DIRECTORY translates into forcing a directory lookup.
1404 */
1405static inline int lookup_flags(unsigned int f)
1406{
1407 unsigned long retval = LOOKUP_FOLLOW;
1408
1409 if (f & O_NOFOLLOW)
1410 retval &= ~LOOKUP_FOLLOW;
1411
1da177e4
LT		 1412 if (f & O_DIRECTORY)
1413 retval |= LOOKUP_DIRECTORY;
1414
1415 return retval;
1416}
1417
1418/*
1419 * p1 and p2 should be directories on the same fs.
1420 */
1421struct dentry *lock_rename(struct dentry *p1, struct dentry *p2)
1422{
1423 struct dentry *p;
1424
1425 if (p1 == p2) {
f2eace23 1426 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1da177e4
LT		 1427 return NULL;
1428 }
1429
a11f3a05 1430 mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex);
1da177e4
LT		 1431
1432 for (p = p1; p->d_parent != p; p = p->d_parent) {
1433 if (p->d_parent == p2) {
f2eace23
IM		 1434 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT);
1435 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD);
1da177e4
LT		 1436 return p;
1437 }
1438 }
1439
1440 for (p = p2; p->d_parent != p; p = p->d_parent) {
1441 if (p->d_parent == p1) {
f2eace23
IM		 1442 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1443 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD);
1da177e4
LT		 1444 return p;
1445 }
1446 }
1447
f2eace23
IM		 1448 mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT);
1449 mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD);
1da177e4
LT		 1450 return NULL;
1451}
1452
1453void unlock_rename(struct dentry *p1, struct dentry *p2)
1454{
1b1dcc1b 1455 mutex_unlock(&p1->d_inode->i_mutex);
1da177e4 1456 if (p1 != p2) {
1b1dcc1b 1457 mutex_unlock(&p2->d_inode->i_mutex);
a11f3a05 1458 mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex);
1da177e4
LT		 1459 }
1460}
1461
1462int vfs_create(struct inode *dir, struct dentry *dentry, int mode,
1463 struct nameidata *nd)
1464{
1465 int error = may_create(dir, dentry, nd);
1466
1467 if (error)
1468 return error;
1469
1470 if (!dir->i_op || !dir->i_op->create)
1471 return -EACCES; /* shouldn't it be ENOSYS? */
1472 mode &= S_IALLUGO;
1473 mode |= S_IFREG;
1474 error = security_inode_create(dir, dentry, mode);
1475 if (error)
1476 return error;
1477 DQUOT_INIT(dir);
1478 error = dir->i_op->create(dir, dentry, mode, nd);
a74574aa 1479 if (!error)
f38aa942 1480 fsnotify_create(dir, dentry);
1da177e4
LT		 1481 return error;
1482}
1483
1484int may_open(struct nameidata *nd, int acc_mode, int flag)
1485{
1486 struct dentry *dentry = nd->dentry;
1487 struct inode *inode = dentry->d_inode;
1488 int error;
1489
1490 if (!inode)
1491 return -ENOENT;
1492
1493 if (S_ISLNK(inode->i_mode))
1494 return -ELOOP;
1495
1496 if (S_ISDIR(inode->i_mode) && (flag & FMODE_WRITE))
1497 return -EISDIR;
1498
e4543edd 1499 error = vfs_permission(nd, acc_mode);
1da177e4
LT		 1500 if (error)
1501 return error;
1502
1503 /*
1504 * FIFO's, sockets and device files are special: they don't
1505 * actually live on the filesystem itself, and as such you
1506 * can write to them even if the filesystem is read-only.
1507 */
1508 if (S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
1509 flag &= ~O_TRUNC;
1510 } else if (S_ISBLK(inode->i_mode) || S_ISCHR(inode->i_mode)) {
1511 if (nd->mnt->mnt_flags & MNT_NODEV)
1512 return -EACCES;
1513
1514 flag &= ~O_TRUNC;
1515 } else if (IS_RDONLY(inode) && (flag & FMODE_WRITE))
1516 return -EROFS;
1517 /*
1518 * An append-only file must be opened in append mode for writing.
1519 */
1520 if (IS_APPEND(inode)) {
1521 if ((flag & FMODE_WRITE) && !(flag & O_APPEND))
1522 return -EPERM;
1523 if (flag & O_TRUNC)
1524 return -EPERM;
1525 }
1526
1527 /* O_NOATIME can only be set by the owner or superuser */
1528 if (flag & O_NOATIME)
1529 if (current->fsuid != inode->i_uid && !capable(CAP_FOWNER))
1530 return -EPERM;
1531
1532 /*
1533 * Ensure there are no outstanding leases on the file.
1534 */
1535 error = break_lease(inode, flag);
1536 if (error)
1537 return error;
1538
1539 if (flag & O_TRUNC) {
1540 error = get_write_access(inode);
1541 if (error)
1542 return error;
1543
1544 /*
1545 * Refuse to truncate files with mandatory locks held on them.
1546 */
1547 error = locks_verify_locked(inode);
1548 if (!error) {
1549 DQUOT_INIT(inode);
1550
4a30131e 1551 error = do_truncate(dentry, 0, ATTR_MTIME|ATTR_CTIME, NULL);
1da177e4
LT		 1552 }
1553 put_write_access(inode);
1554 if (error)
1555 return error;
1556 } else
1557 if (flag & FMODE_WRITE)
1558 DQUOT_INIT(inode);
1559
1560 return 0;
1561}
1562
1563/*
1564 * open_namei()
1565 *
1566 * namei for open - this is in fact almost the whole open-routine.
1567 *
1568 * Note that the low bits of "flag" aren't the same as in the open
1569 * system call - they are 00 - no permissions needed
1570 * 01 - read permission needed
1571 * 10 - write permission needed
1572 * 11 - read/write permissions needed
1573 * which is a lot more logical, and also allows the "no perm" needed
1574 * for symlinks (where the permissions are checked later).
1575 * SMP-safe
1576 */
5590ff0d
UD		 1577int open_namei(int dfd, const char *pathname, int flag,
1578 int mode, struct nameidata *nd)
1da177e4 1579{
834f2a4a 1580 int acc_mode, error;
4e7506e4 1581 struct path path;
1da177e4
LT		 1582 struct dentry *dir;
1583 int count = 0;
1584
1585 acc_mode = ACC_MODE(flag);
1586
834f2a4a
TM		 1587 /* O_TRUNC implies we need access checks for write permissions */
1588 if (flag & O_TRUNC)
1589 acc_mode |= MAY_WRITE;
1590
1da177e4
LT		 1591 /* Allow the LSM permission hook to distinguish append
1592 access from general write access. */
1593 if (flag & O_APPEND)
1594 acc_mode |= MAY_APPEND;
1595
1da177e4
LT		 1596 /*
1597 * The simplest case - just a plain lookup.
1598 */
1599 if (!(flag & O_CREAT)) {
5590ff0d
UD		 1600 error = path_lookup_open(dfd, pathname, lookup_flags(flag),
1601 nd, flag);
1da177e4
LT		 1602 if (error)
1603 return error;
1604 goto ok;
1605 }
1606
1607 /*
1608 * Create - we need to know the parent.
1609 */
5590ff0d 1610 error = path_lookup_create(dfd,pathname,LOOKUP_PARENT,nd,flag,mode);
1da177e4
LT		 1611 if (error)
1612 return error;
1613
1614 /*
1615 * We have the parent and last component. First of all, check
1616 * that we are not asked to creat(2) an obvious directory - that
1617 * will not do.
1618 */
1619 error = -EISDIR;
1620 if (nd->last_type != LAST_NORM || nd->last.name[nd->last.len])
1621 goto exit;
1622
1623 dir = nd->dentry;
1624 nd->flags &= ~LOOKUP_PARENT;
1b1dcc1b 1625 mutex_lock(&dir->d_inode->i_mutex);
49705b77 1626 path.dentry = lookup_hash(nd);
d73ffe16 1627 path.mnt = nd->mnt;
1da177e4
LT		 1628
1629do_last:
4e7506e4
AV		 1630 error = PTR_ERR(path.dentry);
1631 if (IS_ERR(path.dentry)) {
1b1dcc1b 1632 mutex_unlock(&dir->d_inode->i_mutex);
1da177e4
LT		 1633 goto exit;
1634 }
1635
4af4c52f
OD		 1636 if (IS_ERR(nd->intent.open.file)) {
1637 mutex_unlock(&dir->d_inode->i_mutex);
1638 error = PTR_ERR(nd->intent.open.file);
1639 goto exit_dput;
1640 }
1641
1da177e4 1642 /* Negative dentry, just create the file */
4e7506e4 1643 if (!path.dentry->d_inode) {
1da177e4
LT		 1644 if (!IS_POSIXACL(dir->d_inode))
1645 mode &= ~current->fs->umask;
4e7506e4 1646 error = vfs_create(dir->d_inode, path.dentry, mode, nd);
1b1dcc1b 1647 mutex_unlock(&dir->d_inode->i_mutex);
1da177e4 1648 dput(nd->dentry);
4e7506e4 1649 nd->dentry = path.dentry;
1da177e4
LT		 1650 if (error)
1651 goto exit;
1652 /* Don't check for write permission, don't truncate */
1653 acc_mode = 0;
1654 flag &= ~O_TRUNC;
1655 goto ok;
1656 }
1657
1658 /*
1659 * It already exists.
1660 */
1b1dcc1b 1661 mutex_unlock(&dir->d_inode->i_mutex);
3e2efce0 1662 audit_inode_update(path.dentry->d_inode);
1da177e4
LT		 1663
1664 error = -EEXIST;
1665 if (flag & O_EXCL)
1666 goto exit_dput;
1667
e13b210f 1668 if (__follow_mount(&path)) {
1da177e4 1669 error = -ELOOP;
ba7a4c1a
AV		 1670 if (flag & O_NOFOLLOW)
1671 goto exit_dput;
1da177e4 1672 }
3e2efce0 1673
1da177e4 1674 error = -ENOENT;
4e7506e4 1675 if (!path.dentry->d_inode)
1da177e4 1676 goto exit_dput;
4e7506e4 1677 if (path.dentry->d_inode->i_op && path.dentry->d_inode->i_op->follow_link)
1da177e4
LT		 1678 goto do_link;
1679
09dd17d3 1680 path_to_nameidata(&path, nd);
1da177e4 1681 error = -EISDIR;
4e7506e4 1682 if (path.dentry->d_inode && S_ISDIR(path.dentry->d_inode->i_mode))
1da177e4
LT		 1683 goto exit;
1684ok:
1685 error = may_open(nd, acc_mode, flag);
1686 if (error)
1687 goto exit;
1688 return 0;
1689
1690exit_dput:
09dd17d3 1691 dput_path(&path, nd);
1da177e4 1692exit:
834f2a4a
TM		 1693 if (!IS_ERR(nd->intent.open.file))
1694 release_open_intent(nd);
1da177e4
LT		 1695 path_release(nd);
1696 return error;
1697
1698do_link:
1699 error = -ELOOP;
1700 if (flag & O_NOFOLLOW)
1701 goto exit_dput;
1702 /*
1703 * This is subtle. Instead of calling do_follow_link() we do the
1704 * thing by hands. The reason is that this way we have zero link_count
1705 * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT.
1706 * After that we have the parent and last component, i.e.
1707 * we are in the same situation as after the first path_walk().
1708 * Well, almost - if the last component is normal we get its copy
1709 * stored in nd->last.name and we will have to putname() it when we
1710 * are done. Procfs-like symlinks just set LAST_BIND.
1711 */
1712 nd->flags |= LOOKUP_PARENT;
4e7506e4 1713 error = security_inode_follow_link(path.dentry, nd);
1da177e4
LT		 1714 if (error)
1715 goto exit_dput;
cd4e91d3 1716 error = __do_follow_link(&path, nd);
de459215
KK		 1717 if (error) {
1718 /* Does someone understand code flow here? Or it is only
1719 * me so stupid? Anathema to whoever designed this non-sense
1720 * with "intent.open".
1721 */
1722 release_open_intent(nd);
1da177e4 1723 return error;
de459215 1724 }
1da177e4 1725 nd->flags &= ~LOOKUP_PARENT;
d671d5e5 1726 if (nd->last_type == LAST_BIND)
1da177e4 1727 goto ok;
1da177e4
LT		 1728 error = -EISDIR;
1729 if (nd->last_type != LAST_NORM)
1730 goto exit;
1731 if (nd->last.name[nd->last.len]) {
82984114 1732 __putname(nd->last.name);
1da177e4
LT		 1733 goto exit;
1734 }
1735 error = -ELOOP;
1736 if (count++==32) {
82984114 1737 __putname(nd->last.name);
1da177e4
LT		 1738 goto exit;
1739 }
1740 dir = nd->dentry;
1b1dcc1b 1741 mutex_lock(&dir->d_inode->i_mutex);
49705b77 1742 path.dentry = lookup_hash(nd);
d671d5e5 1743 path.mnt = nd->mnt;
82984114 1744 __putname(nd->last.name);
1da177e4
LT		 1745 goto do_last;
1746}
1747
1748/**
1749 * lookup_create - lookup a dentry, creating it if it doesn't exist
1750 * @nd: nameidata info
1751 * @is_dir: directory flag
1752 *
1753 * Simple function to lookup and return a dentry and create it
1754 * if it doesn't exist. Is SMP-safe.
c663e5d8 1755 *
1b1dcc1b 1756 * Returns with nd->dentry->d_inode->i_mutex locked.
1da177e4
LT		 1757 */
1758struct dentry *lookup_create(struct nameidata *nd, int is_dir)
1759{
c663e5d8 1760 struct dentry *dentry = ERR_PTR(-EEXIST);
1da177e4 1761
f2eace23 1762 mutex_lock_nested(&nd->dentry->d_inode->i_mutex, I_MUTEX_PARENT);
c663e5d8
CH		 1763 /*
1764 * Yucky last component or no last component at all?
1765 * (foo/., foo/.., /////)
1766 */
1da177e4
LT		 1767 if (nd->last_type != LAST_NORM)
1768 goto fail;
1769 nd->flags &= ~LOOKUP_PARENT;
c663e5d8
CH		 1770
1771 /*
1772 * Do the final lookup.
1773 */
49705b77 1774 dentry = lookup_hash(nd);
1da177e4
LT		 1775 if (IS_ERR(dentry))
1776 goto fail;
c663e5d8
CH		 1777
1778 /*
1779 * Special case - lookup gave negative, but... we had foo/bar/
1780 * From the vfs_mknod() POV we just have a negative dentry -
1781 * all is fine. Let's be bastards - you had / on the end, you've
1782 * been asking for (non-existent) directory. -ENOENT for you.
1783 */
1da177e4
LT		 1784 if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
1785 goto enoent;
1786 return dentry;
1787enoent:
1788 dput(dentry);
1789 dentry = ERR_PTR(-ENOENT);
1790fail:
1791 return dentry;
1792}
f81a0bff 1793EXPORT_SYMBOL_GPL(lookup_create);
1da177e4
LT		 1794
1795int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
1796{
1797 int error = may_create(dir, dentry, NULL);
1798
1799 if (error)
1800 return error;
1801
1802 if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD))
1803 return -EPERM;
1804
1805 if (!dir->i_op || !dir->i_op->mknod)
1806 return -EPERM;
1807
1808 error = security_inode_mknod(dir, dentry, mode, dev);
1809 if (error)
1810 return error;
1811
1812 DQUOT_INIT(dir);
1813 error = dir->i_op->mknod(dir, dentry, mode, dev);
a74574aa 1814 if (!error)
f38aa942 1815 fsnotify_create(dir, dentry);
1da177e4
LT		 1816 return error;
1817}
1818
5590ff0d
UD		 1819asmlinkage long sys_mknodat(int dfd, const char __user *filename, int mode,
1820 unsigned dev)
1da177e4
LT		 1821{
1822 int error = 0;
1823 char * tmp;
1824 struct dentry * dentry;
1825 struct nameidata nd;
1826
1827 if (S_ISDIR(mode))
1828 return -EPERM;
1829 tmp = getname(filename);
1830 if (IS_ERR(tmp))
1831 return PTR_ERR(tmp);
1832
5590ff0d 1833 error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd);
1da177e4
LT		 1834 if (error)
1835 goto out;
1836 dentry = lookup_create(&nd, 0);
1837 error = PTR_ERR(dentry);
1838
1839 if (!IS_POSIXACL(nd.dentry->d_inode))
1840 mode &= ~current->fs->umask;
1841 if (!IS_ERR(dentry)) {
1842 switch (mode & S_IFMT) {
1843 case 0: case S_IFREG:
1844 error = vfs_create(nd.dentry->d_inode,dentry,mode,&nd);
1845 break;
1846 case S_IFCHR: case S_IFBLK:
1847 error = vfs_mknod(nd.dentry->d_inode,dentry,mode,
1848 new_decode_dev(dev));
1849 break;
1850 case S_IFIFO: case S_IFSOCK:
1851 error = vfs_mknod(nd.dentry->d_inode,dentry,mode,0);
1852 break;
1853 case S_IFDIR:
1854 error = -EPERM;
1855 break;
1856 default:
1857 error = -EINVAL;
1858 }
1859 dput(dentry);
1860 }
1b1dcc1b 1861 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 1862 path_release(&nd);
1863out:
1864 putname(tmp);
1865
1866 return error;
1867}
1868
5590ff0d
UD		 1869asmlinkage long sys_mknod(const char __user *filename, int mode, unsigned dev)
1870{
1871 return sys_mknodat(AT_FDCWD, filename, mode, dev);
1872}
1873
1da177e4
LT		 1874int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1875{
1876 int error = may_create(dir, dentry, NULL);
1877
1878 if (error)
1879 return error;
1880
1881 if (!dir->i_op || !dir->i_op->mkdir)
1882 return -EPERM;
1883
1884 mode &= (S_IRWXUGO|S_ISVTX);
1885 error = security_inode_mkdir(dir, dentry, mode);
1886 if (error)
1887 return error;
1888
1889 DQUOT_INIT(dir);
1890 error = dir->i_op->mkdir(dir, dentry, mode);
a74574aa 1891 if (!error)
f38aa942 1892 fsnotify_mkdir(dir, dentry);
1da177e4
LT		 1893 return error;
1894}
1895
5590ff0d 1896asmlinkage long sys_mkdirat(int dfd, const char __user *pathname, int mode)
1da177e4
LT		 1897{
1898 int error = 0;
1899 char * tmp;
1900
1901 tmp = getname(pathname);
1902 error = PTR_ERR(tmp);
1903 if (!IS_ERR(tmp)) {
1904 struct dentry *dentry;
1905 struct nameidata nd;
1906
5590ff0d 1907 error = do_path_lookup(dfd, tmp, LOOKUP_PARENT, &nd);
1da177e4
LT		 1908 if (error)
1909 goto out;
1910 dentry = lookup_create(&nd, 1);
1911 error = PTR_ERR(dentry);
1912 if (!IS_ERR(dentry)) {
1913 if (!IS_POSIXACL(nd.dentry->d_inode))
1914 mode &= ~current->fs->umask;
1915 error = vfs_mkdir(nd.dentry->d_inode, dentry, mode);
1916 dput(dentry);
1917 }
1b1dcc1b 1918 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 1919 path_release(&nd);
1920out:
1921 putname(tmp);
1922 }
1923
1924 return error;
1925}
1926
5590ff0d
UD		 1927asmlinkage long sys_mkdir(const char __user *pathname, int mode)
1928{
1929 return sys_mkdirat(AT_FDCWD, pathname, mode);
1930}
1931
1da177e4
LT		 1932/*
1933 * We try to drop the dentry early: we should have
1934 * a usage count of 2 if we're the only user of this
1935 * dentry, and if that is true (possibly after pruning
1936 * the dcache), then we drop the dentry now.
1937 *
1938 * A low-level filesystem can, if it choses, legally
1939 * do a
1940 *
1941 * if (!d_unhashed(dentry))
1942 * return -EBUSY;
1943 *
1944 * if it cannot handle the case of removing a directory
1945 * that is still in use by something else..
1946 */
1947void dentry_unhash(struct dentry *dentry)
1948{
1949 dget(dentry);
1950 if (atomic_read(&dentry->d_count))
1951 shrink_dcache_parent(dentry);
1952 spin_lock(&dcache_lock);
1953 spin_lock(&dentry->d_lock);
1954 if (atomic_read(&dentry->d_count) == 2)
1955 __d_drop(dentry);
1956 spin_unlock(&dentry->d_lock);
1957 spin_unlock(&dcache_lock);
1958}
1959
1960int vfs_rmdir(struct inode *dir, struct dentry *dentry)
1961{
1962 int error = may_delete(dir, dentry, 1);
1963
1964 if (error)
1965 return error;
1966
1967 if (!dir->i_op || !dir->i_op->rmdir)
1968 return -EPERM;
1969
1970 DQUOT_INIT(dir);
1971
1b1dcc1b 1972 mutex_lock(&dentry->d_inode->i_mutex);
1da177e4
LT		 1973 dentry_unhash(dentry);
1974 if (d_mountpoint(dentry))
1975 error = -EBUSY;
1976 else {
1977 error = security_inode_rmdir(dir, dentry);
1978 if (!error) {
1979 error = dir->i_op->rmdir(dir, dentry);
1980 if (!error)
1981 dentry->d_inode->i_flags |= S_DEAD;
1982 }
1983 }
1b1dcc1b 1984 mutex_unlock(&dentry->d_inode->i_mutex);
1da177e4 1985 if (!error) {
1da177e4
LT		 1986 d_delete(dentry);
1987 }
1988 dput(dentry);
1989
1990 return error;
1991}
1992
5590ff0d 1993static long do_rmdir(int dfd, const char __user *pathname)
1da177e4
LT		 1994{
1995 int error = 0;
1996 char * name;
1997 struct dentry *dentry;
1998 struct nameidata nd;
1999
2000 name = getname(pathname);
2001 if(IS_ERR(name))
2002 return PTR_ERR(name);
2003
5590ff0d 2004 error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd);
1da177e4
LT		 2005 if (error)
2006 goto exit;
2007
2008 switch(nd.last_type) {
2009 case LAST_DOTDOT:
2010 error = -ENOTEMPTY;
2011 goto exit1;
2012 case LAST_DOT:
2013 error = -EINVAL;
2014 goto exit1;
2015 case LAST_ROOT:
2016 error = -EBUSY;
2017 goto exit1;
2018 }
f2eace23 2019 mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT);
49705b77 2020 dentry = lookup_hash(&nd);
1da177e4
LT		 2021 error = PTR_ERR(dentry);
2022 if (!IS_ERR(dentry)) {
2023 error = vfs_rmdir(nd.dentry->d_inode, dentry);
2024 dput(dentry);
2025 }
1b1dcc1b 2026 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2027exit1:
2028 path_release(&nd);
2029exit:
2030 putname(name);
2031 return error;
2032}
2033
5590ff0d
UD		 2034asmlinkage long sys_rmdir(const char __user *pathname)
2035{
2036 return do_rmdir(AT_FDCWD, pathname);
2037}
2038
1da177e4
LT		 2039int vfs_unlink(struct inode *dir, struct dentry *dentry)
2040{
2041 int error = may_delete(dir, dentry, 0);
2042
2043 if (error)
2044 return error;
2045
2046 if (!dir->i_op || !dir->i_op->unlink)
2047 return -EPERM;
2048
2049 DQUOT_INIT(dir);
2050
1b1dcc1b 2051 mutex_lock(&dentry->d_inode->i_mutex);
1da177e4
LT		 2052 if (d_mountpoint(dentry))
2053 error = -EBUSY;
2054 else {
2055 error = security_inode_unlink(dir, dentry);
2056 if (!error)
2057 error = dir->i_op->unlink(dir, dentry);
2058 }
1b1dcc1b 2059 mutex_unlock(&dentry->d_inode->i_mutex);
1da177e4
LT		 2060
2061 /* We don't d_delete() NFS sillyrenamed files--they still exist. */
2062 if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) {
e234f35c 2063 d_delete(dentry);
1da177e4 2064 }
0eeca283 2065
1da177e4
LT		 2066 return error;
2067}
2068
2069/*
2070 * Make sure that the actual truncation of the file will occur outside its
1b1dcc1b 2071 * directory's i_mutex. Truncate can take a long time if there is a lot of
1da177e4
LT		 2072 * writeout happening, and we don't want to prevent access to the directory
2073 * while waiting on the I/O.
2074 */
5590ff0d 2075static long do_unlinkat(int dfd, const char __user *pathname)
1da177e4
LT		 2076{
2077 int error = 0;
2078 char * name;
2079 struct dentry *dentry;
2080 struct nameidata nd;
2081 struct inode *inode = NULL;
2082
2083 name = getname(pathname);
2084 if(IS_ERR(name))
2085 return PTR_ERR(name);
2086
5590ff0d 2087 error = do_path_lookup(dfd, name, LOOKUP_PARENT, &nd);
1da177e4
LT		 2088 if (error)
2089 goto exit;
2090 error = -EISDIR;
2091 if (nd.last_type != LAST_NORM)
2092 goto exit1;
f2eace23 2093 mutex_lock_nested(&nd.dentry->d_inode->i_mutex, I_MUTEX_PARENT);
49705b77 2094 dentry = lookup_hash(&nd);
1da177e4
LT		 2095 error = PTR_ERR(dentry);
2096 if (!IS_ERR(dentry)) {
2097 /* Why not before? Because we want correct error value */
2098 if (nd.last.name[nd.last.len])
2099 goto slashes;
2100 inode = dentry->d_inode;
2101 if (inode)
2102 atomic_inc(&inode->i_count);
2103 error = vfs_unlink(nd.dentry->d_inode, dentry);
2104 exit2:
2105 dput(dentry);
2106 }
1b1dcc1b 2107 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2108 if (inode)
2109 iput(inode); /* truncate the inode here */
2110exit1:
2111 path_release(&nd);
2112exit:
2113 putname(name);
2114 return error;
2115
2116slashes:
2117 error = !dentry->d_inode ? -ENOENT :
2118 S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR;
2119 goto exit2;
2120}
2121
5590ff0d
UD		 2122asmlinkage long sys_unlinkat(int dfd, const char __user *pathname, int flag)
2123{
2124 if ((flag & ~AT_REMOVEDIR) != 0)
2125 return -EINVAL;
2126
2127 if (flag & AT_REMOVEDIR)
2128 return do_rmdir(dfd, pathname);
2129
2130 return do_unlinkat(dfd, pathname);
2131}
2132
2133asmlinkage long sys_unlink(const char __user *pathname)
2134{
2135 return do_unlinkat(AT_FDCWD, pathname);
2136}
2137
1da177e4
LT		 2138int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname, int mode)
2139{
2140 int error = may_create(dir, dentry, NULL);
2141
2142 if (error)
2143 return error;
2144
2145 if (!dir->i_op || !dir->i_op->symlink)
2146 return -EPERM;
2147
2148 error = security_inode_symlink(dir, dentry, oldname);
2149 if (error)
2150 return error;
2151
2152 DQUOT_INIT(dir);
2153 error = dir->i_op->symlink(dir, dentry, oldname);
a74574aa 2154 if (!error)
f38aa942 2155 fsnotify_create(dir, dentry);
1da177e4
LT		 2156 return error;
2157}
2158
5590ff0d
UD		 2159asmlinkage long sys_symlinkat(const char __user *oldname,
2160 int newdfd, const char __user *newname)
1da177e4
LT		 2161{
2162 int error = 0;
2163 char * from;
2164 char * to;
2165
2166 from = getname(oldname);
2167 if(IS_ERR(from))
2168 return PTR_ERR(from);
2169 to = getname(newname);
2170 error = PTR_ERR(to);
2171 if (!IS_ERR(to)) {
2172 struct dentry *dentry;
2173 struct nameidata nd;
2174
5590ff0d 2175 error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd);
1da177e4
LT		 2176 if (error)
2177 goto out;
2178 dentry = lookup_create(&nd, 0);
2179 error = PTR_ERR(dentry);
2180 if (!IS_ERR(dentry)) {
2181 error = vfs_symlink(nd.dentry->d_inode, dentry, from, S_IALLUGO);
2182 dput(dentry);
2183 }
1b1dcc1b 2184 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2185 path_release(&nd);
2186out:
2187 putname(to);
2188 }
2189 putname(from);
2190 return error;
2191}
2192
5590ff0d
UD		 2193asmlinkage long sys_symlink(const char __user *oldname, const char __user *newname)
2194{
2195 return sys_symlinkat(oldname, AT_FDCWD, newname);
2196}
2197
1da177e4
LT		 2198int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry)
2199{
2200 struct inode *inode = old_dentry->d_inode;
2201 int error;
2202
2203 if (!inode)
2204 return -ENOENT;
2205
2206 error = may_create(dir, new_dentry, NULL);
2207 if (error)
2208 return error;
2209
2210 if (dir->i_sb != inode->i_sb)
2211 return -EXDEV;
2212
2213 /*
2214 * A link to an append-only or immutable file cannot be created.
2215 */
2216 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2217 return -EPERM;
2218 if (!dir->i_op || !dir->i_op->link)
2219 return -EPERM;
2220 if (S_ISDIR(old_dentry->d_inode->i_mode))
2221 return -EPERM;
2222
2223 error = security_inode_link(old_dentry, dir, new_dentry);
2224 if (error)
2225 return error;
2226
1b1dcc1b 2227 mutex_lock(&old_dentry->d_inode->i_mutex);
1da177e4
LT		 2228 DQUOT_INIT(dir);
2229 error = dir->i_op->link(old_dentry, dir, new_dentry);
1b1dcc1b 2230 mutex_unlock(&old_dentry->d_inode->i_mutex);
e31e14ec 2231 if (!error)
f38aa942 2232 fsnotify_create(dir, new_dentry);
1da177e4
LT		 2233 return error;
2234}
2235
2236/*
2237 * Hardlinks are often used in delicate situations. We avoid
2238 * security-related surprises by not following symlinks on the
2239 * newname. --KAB
2240 *
2241 * We don't follow them on the oldname either to be compatible
2242 * with linux 2.0, and to avoid hard-linking to directories
2243 * and other special files. --ADM
2244 */
5590ff0d 2245asmlinkage long sys_linkat(int olddfd, const char __user *oldname,
c04030e1
UD		 2246 int newdfd, const char __user *newname,
2247 int flags)
1da177e4
LT		 2248{
2249 struct dentry *new_dentry;
2250 struct nameidata nd, old_nd;
2251 int error;
2252 char * to;
2253
45c9b11a 2254 if ((flags & ~AT_SYMLINK_FOLLOW) != 0)
c04030e1
UD		 2255 return -EINVAL;
2256
1da177e4
LT		 2257 to = getname(newname);
2258 if (IS_ERR(to))
2259 return PTR_ERR(to);
2260
45c9b11a
UD		 2261 error = __user_walk_fd(olddfd, oldname,
2262 flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0,
2263 &old_nd);
1da177e4
LT		 2264 if (error)
2265 goto exit;
5590ff0d 2266 error = do_path_lookup(newdfd, to, LOOKUP_PARENT, &nd);
1da177e4
LT		 2267 if (error)
2268 goto out;
2269 error = -EXDEV;
2270 if (old_nd.mnt != nd.mnt)
2271 goto out_release;
2272 new_dentry = lookup_create(&nd, 0);
2273 error = PTR_ERR(new_dentry);
2274 if (!IS_ERR(new_dentry)) {
2275 error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry);
2276 dput(new_dentry);
2277 }
1b1dcc1b 2278 mutex_unlock(&nd.dentry->d_inode->i_mutex);
1da177e4
LT		 2279out_release:
2280 path_release(&nd);
2281out:
2282 path_release(&old_nd);
2283exit:
2284 putname(to);
2285
2286 return error;
2287}
2288
5590ff0d
UD		 2289asmlinkage long sys_link(const char __user *oldname, const char __user *newname)
2290{
c04030e1 2291 return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0);
5590ff0d
UD		 2292}
2293
1da177e4
LT		 2294/*
2295 * The worst of all namespace operations - renaming directory. "Perverted"
2296 * doesn't even start to describe it. Somebody in UCB had a heck of a trip...
2297 * Problems:
2298 * a) we can get into loop creation. Check is done in is_subdir().
2299 * b) race potential - two innocent renames can create a loop together.
2300 * That's where 4.4 screws up. Current fix: serialization on
a11f3a05 2301 * sb->s_vfs_rename_mutex. We might be more accurate, but that's another
1da177e4
LT		 2302 * story.
2303 * c) we have to lock _three_ objects - parents and victim (if it exists).
1b1dcc1b 2304 * And that - after we got ->i_mutex on parents (until then we don't know
1da177e4
LT		 2305 * whether the target exists). Solution: try to be smart with locking
2306 * order for inodes. We rely on the fact that tree topology may change
a11f3a05 2307 * only under ->s_vfs_rename_mutex _and_ that parent of the object we
1da177e4
LT		 2308 * move will be locked. Thus we can rank directories by the tree
2309 * (ancestors first) and rank all non-directories after them.
2310 * That works since everybody except rename does "lock parent, lookup,
a11f3a05 2311 * lock child" and rename is under ->s_vfs_rename_mutex.
1da177e4
LT		 2312 * HOWEVER, it relies on the assumption that any object with ->lookup()
2313 * has no more than 1 dentry. If "hybrid" objects will ever appear,
2314 * we'd better make sure that there's no link(2) for them.
2315 * d) some filesystems don't support opened-but-unlinked directories,
2316 * either because of layout or because they are not ready to deal with
2317 * all cases correctly. The latter will be fixed (taking this sort of
2318 * stuff into VFS), but the former is not going away. Solution: the same
2319 * trick as in rmdir().
2320 * e) conversion from fhandle to dentry may come in the wrong moment - when
1b1dcc1b 2321 * we are removing the target. Solution: we will have to grab ->i_mutex
1da177e4 2322 * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on
1b1dcc1b 2323 * ->i_mutex on parents, which works but leads to some truely excessive
1da177e4
LT		 2324 * locking].
2325 */
75c96f85
AB		 2326static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry,
2327 struct inode *new_dir, struct dentry *new_dentry)
1da177e4
LT		 2328{
2329 int error = 0;
2330 struct inode *target;
2331
2332 /*
2333 * If we are going to change the parent - check write permissions,
2334 * we'll need to flip '..'.
2335 */
2336 if (new_dir != old_dir) {
2337 error = permission(old_dentry->d_inode, MAY_WRITE, NULL);
2338 if (error)
2339 return error;
2340 }
2341
2342 error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry);
2343 if (error)
2344 return error;
2345
2346 target = new_dentry->d_inode;
2347 if (target) {
1b1dcc1b 2348 mutex_lock(&target->i_mutex);
1da177e4
LT		 2349 dentry_unhash(new_dentry);
2350 }
2351 if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry))
2352 error = -EBUSY;
2353 else
2354 error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry);
2355 if (target) {
2356 if (!error)
2357 target->i_flags |= S_DEAD;
1b1dcc1b 2358 mutex_unlock(&target->i_mutex);
1da177e4
LT		 2359 if (d_unhashed(new_dentry))
2360 d_rehash(new_dentry);
2361 dput(new_dentry);
2362 }
e31e14ec 2363 if (!error)
1da177e4 2364 d_move(old_dentry,new_dentry);
1da177e4
LT		 2365 return error;
2366}
2367
75c96f85
AB		 2368static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry,
2369 struct inode *new_dir, struct dentry *new_dentry)
1da177e4
LT		 2370{
2371 struct inode *target;
2372 int error;
2373
2374 error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry);
2375 if (error)
2376 return error;
2377
2378 dget(new_dentry);
2379 target = new_dentry->d_inode;
2380 if (target)
1b1dcc1b 2381 mutex_lock(&target->i_mutex);
1da177e4
LT		 2382 if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry))
2383 error = -EBUSY;
2384 else
2385 error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry);
2386 if (!error) {
2387 /* The following d_move() should become unconditional */
2388 if (!(old_dir->i_sb->s_type->fs_flags & FS_ODD_RENAME))
2389 d_move(old_dentry, new_dentry);
1da177e4
LT		 2390 }
2391 if (target)
1b1dcc1b 2392 mutex_unlock(&target->i_mutex);
1da177e4
LT		 2393 dput(new_dentry);
2394 return error;
2395}
2396
2397int vfs_rename(struct inode *old_dir, struct dentry *old_dentry,
2398 struct inode *new_dir, struct dentry *new_dentry)
2399{
2400 int error;
2401 int is_dir = S_ISDIR(old_dentry->d_inode->i_mode);
0eeca283 2402 const char *old_name;
1da177e4
LT		 2403
2404 if (old_dentry->d_inode == new_dentry->d_inode)
2405 return 0;
2406
2407 error = may_delete(old_dir, old_dentry, is_dir);
2408 if (error)
2409 return error;
2410
2411 if (!new_dentry->d_inode)
2412 error = may_create(new_dir, new_dentry, NULL);
2413 else
2414 error = may_delete(new_dir, new_dentry, is_dir);
2415 if (error)
2416 return error;
2417
2418 if (!old_dir->i_op || !old_dir->i_op->rename)
2419 return -EPERM;
2420
2421 DQUOT_INIT(old_dir);
2422 DQUOT_INIT(new_dir);
2423
0eeca283
RL		 2424 old_name = fsnotify_oldname_init(old_dentry->d_name.name);
2425
1da177e4
LT		 2426 if (is_dir)
2427 error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry);
2428 else
2429 error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry);
2430 if (!error) {
0eeca283 2431 const char *new_name = old_dentry->d_name.name;
89204c40
JM		 2432 fsnotify_move(old_dir, new_dir, old_name, new_name, is_dir,
2433 new_dentry->d_inode, old_dentry->d_inode);
1da177e4 2434 }
0eeca283
RL		 2435 fsnotify_oldname_free(old_name);
2436
1da177e4
LT		 2437 return error;
2438}
2439
5590ff0d
UD		 2440static int do_rename(int olddfd, const char *oldname,
2441 int newdfd, const char *newname)
1da177e4
LT		 2442{
2443 int error = 0;
2444 struct dentry * old_dir, * new_dir;
2445 struct dentry * old_dentry, *new_dentry;
2446 struct dentry * trap;
2447 struct nameidata oldnd, newnd;
2448
5590ff0d 2449 error = do_path_lookup(olddfd, oldname, LOOKUP_PARENT, &oldnd);
1da177e4
LT		 2450 if (error)
2451 goto exit;
2452
5590ff0d 2453 error = do_path_lookup(newdfd, newname, LOOKUP_PARENT, &newnd);
1da177e4
LT		 2454 if (error)
2455 goto exit1;
2456
2457 error = -EXDEV;
2458 if (oldnd.mnt != newnd.mnt)
2459 goto exit2;
2460
2461 old_dir = oldnd.dentry;
2462 error = -EBUSY;
2463 if (oldnd.last_type != LAST_NORM)
2464 goto exit2;
2465
2466 new_dir = newnd.dentry;
2467 if (newnd.last_type != LAST_NORM)
2468 goto exit2;
2469
2470 trap = lock_rename(new_dir, old_dir);
2471
49705b77 2472 old_dentry = lookup_hash(&oldnd);
1da177e4
LT		 2473 error = PTR_ERR(old_dentry);
2474 if (IS_ERR(old_dentry))
2475 goto exit3;
2476 /* source must exist */
2477 error = -ENOENT;
2478 if (!old_dentry->d_inode)
2479 goto exit4;
2480 /* unless the source is a directory trailing slashes give -ENOTDIR */
2481 if (!S_ISDIR(old_dentry->d_inode->i_mode)) {
2482 error = -ENOTDIR;
2483 if (oldnd.last.name[oldnd.last.len])
2484 goto exit4;
2485 if (newnd.last.name[newnd.last.len])
2486 goto exit4;
2487 }
2488 /* source should not be ancestor of target */
2489 error = -EINVAL;
2490 if (old_dentry == trap)
2491 goto exit4;
49705b77 2492 new_dentry = lookup_hash(&newnd);
1da177e4
LT		 2493 error = PTR_ERR(new_dentry);
2494 if (IS_ERR(new_dentry))
2495 goto exit4;
2496 /* target should not be an ancestor of source */
2497 error = -ENOTEMPTY;
2498 if (new_dentry == trap)
2499 goto exit5;
2500
2501 error = vfs_rename(old_dir->d_inode, old_dentry,
2502 new_dir->d_inode, new_dentry);
2503exit5:
2504 dput(new_dentry);
2505exit4:
2506 dput(old_dentry);
2507exit3:
2508 unlock_rename(new_dir, old_dir);
2509exit2:
2510 path_release(&newnd);
2511exit1:
2512 path_release(&oldnd);
2513exit:
2514 return error;
2515}
2516
5590ff0d
UD		 2517asmlinkage long sys_renameat(int olddfd, const char __user *oldname,
2518 int newdfd, const char __user *newname)
1da177e4
LT		 2519{
2520 int error;
2521 char * from;
2522 char * to;
2523
2524 from = getname(oldname);
2525 if(IS_ERR(from))
2526 return PTR_ERR(from);
2527 to = getname(newname);
2528 error = PTR_ERR(to);
2529 if (!IS_ERR(to)) {
5590ff0d 2530 error = do_rename(olddfd, from, newdfd, to);
1da177e4
LT		 2531 putname(to);
2532 }
2533 putname(from);
2534 return error;
2535}
2536
5590ff0d
UD		 2537asmlinkage long sys_rename(const char __user *oldname, const char __user *newname)
2538{
2539 return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname);
2540}
2541
1da177e4
LT		 2542int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link)
2543{
2544 int len;
2545
2546 len = PTR_ERR(link);
2547 if (IS_ERR(link))
2548 goto out;
2549
2550 len = strlen(link);
2551 if (len > (unsigned) buflen)
2552 len = buflen;
2553 if (copy_to_user(buffer, link, len))
2554 len = -EFAULT;
2555out:
2556 return len;
2557}
2558
2559/*
2560 * A helper for ->readlink(). This should be used *ONLY* for symlinks that
2561 * have ->follow_link() touching nd only in nd_set_link(). Using (or not
2562 * using) it for any given inode is up to filesystem.
2563 */
2564int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen)
2565{
2566 struct nameidata nd;
cc314eef
LT		 2567 void *cookie;
2568
1da177e4 2569 nd.depth = 0;
cc314eef
LT		 2570 cookie = dentry->d_inode->i_op->follow_link(dentry, &nd);
2571 if (!IS_ERR(cookie)) {
2572 int res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd));
1da177e4 2573 if (dentry->d_inode->i_op->put_link)
cc314eef
LT		 2574 dentry->d_inode->i_op->put_link(dentry, &nd, cookie);
2575 cookie = ERR_PTR(res);
1da177e4 2576 }
cc314eef 2577 return PTR_ERR(cookie);
1da177e4
LT		 2578}
2579
2580int vfs_follow_link(struct nameidata *nd, const char *link)
2581{
2582 return __vfs_follow_link(nd, link);
2583}
2584
2585/* get the link contents into pagecache */
2586static char *page_getlink(struct dentry * dentry, struct page **ppage)
2587{
2588 struct page * page;
2589 struct address_space *mapping = dentry->d_inode->i_mapping;
090d2b18 2590 page = read_mapping_page(mapping, 0, NULL);
1da177e4
LT		 2591 if (IS_ERR(page))
2592 goto sync_fail;
2593 wait_on_page_locked(page);
2594 if (!PageUptodate(page))
2595 goto async_fail;
2596 *ppage = page;
2597 return kmap(page);
2598
2599async_fail:
2600 page_cache_release(page);
2601 return ERR_PTR(-EIO);
2602
2603sync_fail:
2604 return (char*)page;
2605}
2606
2607int page_readlink(struct dentry *dentry, char __user *buffer, int buflen)
2608{
2609 struct page *page = NULL;
2610 char *s = page_getlink(dentry, &page);
2611 int res = vfs_readlink(dentry,buffer,buflen,s);
2612 if (page) {
2613 kunmap(page);
2614 page_cache_release(page);
2615 }
2616 return res;
2617}
2618
cc314eef 2619void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd)
1da177e4 2620{
cc314eef 2621 struct page *page = NULL;
1da177e4 2622 nd_set_link(nd, page_getlink(dentry, &page));
cc314eef 2623 return page;
1da177e4
LT		 2624}
2625
cc314eef 2626void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie)
1da177e4 2627{
cc314eef
LT		 2628 struct page *page = cookie;
2629
2630 if (page) {
1da177e4
LT		 2631 kunmap(page);
2632 page_cache_release(page);
1da177e4
LT		 2633 }
2634}
2635
0adb25d2
KK		 2636int __page_symlink(struct inode *inode, const char *symname, int len,
2637 gfp_t gfp_mask)
1da177e4
LT		 2638{
2639 struct address_space *mapping = inode->i_mapping;
0adb25d2 2640 struct page *page;
1da177e4
LT		 2641 int err = -ENOMEM;
2642 char *kaddr;
2643
7e53cac4 2644retry:
0adb25d2 2645 page = find_or_create_page(mapping, 0, gfp_mask);
1da177e4
LT		 2646 if (!page)
2647 goto fail;
2648 err = mapping->a_ops->prepare_write(NULL, page, 0, len-1);
7e53cac4
N		 2649 if (err == AOP_TRUNCATED_PAGE) {
2650 page_cache_release(page);
2651 goto retry;
2652 }
1da177e4
LT		 2653 if (err)
2654 goto fail_map;
2655 kaddr = kmap_atomic(page, KM_USER0);
2656 memcpy(kaddr, symname, len-1);
2657 kunmap_atomic(kaddr, KM_USER0);
7e53cac4
N		 2658 err = mapping->a_ops->commit_write(NULL, page, 0, len-1);
2659 if (err == AOP_TRUNCATED_PAGE) {
2660 page_cache_release(page);
2661 goto retry;
2662 }
2663 if (err)
2664 goto fail_map;
1da177e4
LT		 2665 /*
2666 * Notice that we are _not_ going to block here - end of page is
2667 * unmapped, so this will only try to map the rest of page, see
2668 * that it is unmapped (typically even will not look into inode -
2669 * ->i_size will be enough for everything) and zero it out.
2670 * OTOH it's obviously correct and should make the page up-to-date.
2671 */
2672 if (!PageUptodate(page)) {
2673 err = mapping->a_ops->readpage(NULL, page);
7e53cac4
N		 2674 if (err != AOP_TRUNCATED_PAGE)
2675 wait_on_page_locked(page);
1da177e4
LT		 2676 } else {
2677 unlock_page(page);
2678 }
2679 page_cache_release(page);
2680 if (err < 0)
2681 goto fail;
2682 mark_inode_dirty(inode);
2683 return 0;
2684fail_map:
2685 unlock_page(page);
2686 page_cache_release(page);
2687fail:
2688 return err;
2689}
2690
0adb25d2
KK		 2691int page_symlink(struct inode *inode, const char *symname, int len)
2692{
2693 return __page_symlink(inode, symname, len,
2694 mapping_gfp_mask(inode->i_mapping));
2695}
2696
1da177e4
LT		 2697struct inode_operations page_symlink_inode_operations = {
2698 .readlink = generic_readlink,
2699 .follow_link = page_follow_link_light,
2700 .put_link = page_put_link,
2701};
2702
2703EXPORT_SYMBOL(__user_walk);
5590ff0d 2704EXPORT_SYMBOL(__user_walk_fd);
1da177e4
LT		 2705EXPORT_SYMBOL(follow_down);
2706EXPORT_SYMBOL(follow_up);
2707EXPORT_SYMBOL(get_write_access); /* binfmt_aout */
2708EXPORT_SYMBOL(getname);
2709EXPORT_SYMBOL(lock_rename);
1da177e4
LT		 2710EXPORT_SYMBOL(lookup_one_len);
2711EXPORT_SYMBOL(page_follow_link_light);
2712EXPORT_SYMBOL(page_put_link);
2713EXPORT_SYMBOL(page_readlink);
0adb25d2 2714EXPORT_SYMBOL(__page_symlink);
1da177e4
LT		 2715EXPORT_SYMBOL(page_symlink);
2716EXPORT_SYMBOL(page_symlink_inode_operations);
2717EXPORT_SYMBOL(path_lookup);
2718EXPORT_SYMBOL(path_release);
2719EXPORT_SYMBOL(path_walk);
2720EXPORT_SYMBOL(permission);
e4543edd 2721EXPORT_SYMBOL(vfs_permission);
8c744fb8 2722EXPORT_SYMBOL(file_permission);
1da177e4
LT		 2723EXPORT_SYMBOL(unlock_rename);
2724EXPORT_SYMBOL(vfs_create);
2725EXPORT_SYMBOL(vfs_follow_link);
2726EXPORT_SYMBOL(vfs_link);
2727EXPORT_SYMBOL(vfs_mkdir);
2728EXPORT_SYMBOL(vfs_mknod);
2729EXPORT_SYMBOL(generic_permission);
2730EXPORT_SYMBOL(vfs_readlink);
2731EXPORT_SYMBOL(vfs_rename);
2732EXPORT_SYMBOL(vfs_rmdir);
2733EXPORT_SYMBOL(vfs_symlink);
2734EXPORT_SYMBOL(vfs_unlink);
2735EXPORT_SYMBOL(dentry_unhash);
2736EXPORT_SYMBOL(generic_readlink);
kernel source for telekom puls (alcatel/mediatek)