]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/libfs.c
namei: simplify the callers of follow_managed()
[mirror_ubuntu-artful-kernel.git] / fs / libfs.c
CommitLineData
1da177e4
LT
1/*
2 * fs/libfs.c
3 * Library for filesystems writers.
4 */
5
ac13a829 6#include <linux/blkdev.h>
630d9c47 7#include <linux/export.h>
1da177e4 8#include <linux/pagemap.h>
5a0e3ad6 9#include <linux/slab.h>
1da177e4
LT
10#include <linux/mount.h>
11#include <linux/vfs.h>
7bb46a67 12#include <linux/quotaops.h>
7cf34c76 13#include <linux/mutex.h>
87dc800b 14#include <linux/namei.h>
2596110a 15#include <linux/exportfs.h>
d5aacad5 16#include <linux/writeback.h>
ff01bb48 17#include <linux/buffer_head.h> /* sync_mapping_buffers */
7cf34c76 18
1da177e4
LT
19#include <asm/uaccess.h>
20
a4464dbc
AV
21#include "internal.h"
22
da502956
NP
23static inline int simple_positive(struct dentry *dentry)
24{
dea655c2 25 return d_really_is_positive(dentry) && !d_unhashed(dentry);
da502956
NP
26}
27
1da177e4
LT
28int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
29 struct kstat *stat)
30{
dea655c2 31 struct inode *inode = d_inode(dentry);
1da177e4
LT
32 generic_fillattr(inode, stat);
33 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
34 return 0;
35}
12f38872 36EXPORT_SYMBOL(simple_getattr);
1da177e4 37
726c3342 38int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
1da177e4 39{
726c3342 40 buf->f_type = dentry->d_sb->s_magic;
1da177e4
LT
41 buf->f_bsize = PAGE_CACHE_SIZE;
42 buf->f_namelen = NAME_MAX;
43 return 0;
44}
12f38872 45EXPORT_SYMBOL(simple_statfs);
1da177e4
LT
46
47/*
48 * Retaining negative dentries for an in-memory filesystem just wastes
49 * memory and lookup time: arrange for them to be deleted immediately.
50 */
b26d4cd3 51int always_delete_dentry(const struct dentry *dentry)
1da177e4
LT
52{
53 return 1;
54}
b26d4cd3
AV
55EXPORT_SYMBOL(always_delete_dentry);
56
57const struct dentry_operations simple_dentry_operations = {
58 .d_delete = always_delete_dentry,
59};
60EXPORT_SYMBOL(simple_dentry_operations);
1da177e4
LT
61
62/*
63 * Lookup the data. This is trivial - if the dentry didn't already
64 * exist, we know it is negative. Set d_op to delete negative dentries.
65 */
00cd8dd3 66struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1da177e4 67{
1da177e4
LT
68 if (dentry->d_name.len > NAME_MAX)
69 return ERR_PTR(-ENAMETOOLONG);
74931da7
AV
70 if (!dentry->d_sb->s_d_op)
71 d_set_d_op(dentry, &simple_dentry_operations);
1da177e4
LT
72 d_add(dentry, NULL);
73 return NULL;
74}
12f38872 75EXPORT_SYMBOL(simple_lookup);
1da177e4 76
1da177e4
LT
77int dcache_dir_open(struct inode *inode, struct file *file)
78{
26fe5750 79 static struct qstr cursor_name = QSTR_INIT(".", 1);
1da177e4 80
0f7fc9e4 81 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
1da177e4
LT
82
83 return file->private_data ? 0 : -ENOMEM;
84}
12f38872 85EXPORT_SYMBOL(dcache_dir_open);
1da177e4
LT
86
87int dcache_dir_close(struct inode *inode, struct file *file)
88{
89 dput(file->private_data);
90 return 0;
91}
12f38872 92EXPORT_SYMBOL(dcache_dir_close);
1da177e4 93
965c8e59 94loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
1da177e4 95{
2fd6b7f5 96 struct dentry *dentry = file->f_path.dentry;
dea655c2 97 mutex_lock(&d_inode(dentry)->i_mutex);
965c8e59 98 switch (whence) {
1da177e4
LT
99 case 1:
100 offset += file->f_pos;
101 case 0:
102 if (offset >= 0)
103 break;
104 default:
dea655c2 105 mutex_unlock(&d_inode(dentry)->i_mutex);
1da177e4
LT
106 return -EINVAL;
107 }
108 if (offset != file->f_pos) {
109 file->f_pos = offset;
110 if (file->f_pos >= 2) {
111 struct list_head *p;
112 struct dentry *cursor = file->private_data;
113 loff_t n = file->f_pos - 2;
114
2fd6b7f5
NP
115 spin_lock(&dentry->d_lock);
116 /* d_lock not required for cursor */
946e51f2 117 list_del(&cursor->d_child);
2fd6b7f5
NP
118 p = dentry->d_subdirs.next;
119 while (n && p != &dentry->d_subdirs) {
1da177e4 120 struct dentry *next;
946e51f2 121 next = list_entry(p, struct dentry, d_child);
2fd6b7f5 122 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
da502956 123 if (simple_positive(next))
1da177e4 124 n--;
da502956 125 spin_unlock(&next->d_lock);
1da177e4
LT
126 p = p->next;
127 }
946e51f2 128 list_add_tail(&cursor->d_child, p);
2fd6b7f5 129 spin_unlock(&dentry->d_lock);
1da177e4
LT
130 }
131 }
dea655c2 132 mutex_unlock(&d_inode(dentry)->i_mutex);
1da177e4
LT
133 return offset;
134}
12f38872 135EXPORT_SYMBOL(dcache_dir_lseek);
1da177e4
LT
136
137/* Relationship between i_mode and the DT_xxx types */
138static inline unsigned char dt_type(struct inode *inode)
139{
140 return (inode->i_mode >> 12) & 15;
141}
142
143/*
144 * Directory is locked and all positive dentries in it are safe, since
145 * for ramfs-type trees they can't go away without unlink() or rmdir(),
146 * both impossible due to the lock on directory.
147 */
148
5f99f4e7 149int dcache_readdir(struct file *file, struct dir_context *ctx)
1da177e4 150{
5f99f4e7
AV
151 struct dentry *dentry = file->f_path.dentry;
152 struct dentry *cursor = file->private_data;
946e51f2 153 struct list_head *p, *q = &cursor->d_child;
1da177e4 154
5f99f4e7
AV
155 if (!dir_emit_dots(file, ctx))
156 return 0;
157 spin_lock(&dentry->d_lock);
158 if (ctx->pos == 2)
159 list_move(q, &dentry->d_subdirs);
160
161 for (p = q->next; p != &dentry->d_subdirs; p = p->next) {
946e51f2 162 struct dentry *next = list_entry(p, struct dentry, d_child);
5f99f4e7
AV
163 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
164 if (!simple_positive(next)) {
165 spin_unlock(&next->d_lock);
166 continue;
167 }
1da177e4 168
5f99f4e7
AV
169 spin_unlock(&next->d_lock);
170 spin_unlock(&dentry->d_lock);
171 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
dea655c2 172 d_inode(next)->i_ino, dt_type(d_inode(next))))
5f99f4e7
AV
173 return 0;
174 spin_lock(&dentry->d_lock);
175 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
176 /* next is still alive */
177 list_move(q, p);
178 spin_unlock(&next->d_lock);
179 p = q;
180 ctx->pos++;
1da177e4 181 }
5f99f4e7 182 spin_unlock(&dentry->d_lock);
1da177e4
LT
183 return 0;
184}
12f38872 185EXPORT_SYMBOL(dcache_readdir);
1da177e4
LT
186
187ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
188{
189 return -EISDIR;
190}
12f38872 191EXPORT_SYMBOL(generic_read_dir);
1da177e4 192
4b6f5d20 193const struct file_operations simple_dir_operations = {
1da177e4
LT
194 .open = dcache_dir_open,
195 .release = dcache_dir_close,
196 .llseek = dcache_dir_lseek,
197 .read = generic_read_dir,
5f99f4e7 198 .iterate = dcache_readdir,
1b061d92 199 .fsync = noop_fsync,
1da177e4 200};
12f38872 201EXPORT_SYMBOL(simple_dir_operations);
1da177e4 202
92e1d5be 203const struct inode_operations simple_dir_inode_operations = {
1da177e4
LT
204 .lookup = simple_lookup,
205};
12f38872 206EXPORT_SYMBOL(simple_dir_inode_operations);
1da177e4 207
759b9775
HD
208static const struct super_operations simple_super_operations = {
209 .statfs = simple_statfs,
210};
211
1da177e4
LT
212/*
213 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
214 * will never be mountable)
215 */
51139ada 216struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
c74a1cbb
AV
217 const struct super_operations *ops,
218 const struct dentry_operations *dops, unsigned long magic)
1da177e4 219{
9249e17f 220 struct super_block *s;
1da177e4
LT
221 struct dentry *dentry;
222 struct inode *root;
26fe5750 223 struct qstr d_name = QSTR_INIT(name, strlen(name));
1da177e4 224
9249e17f 225 s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
1da177e4 226 if (IS_ERR(s))
51139ada 227 return ERR_CAST(s);
1da177e4 228
89a4eb4b 229 s->s_maxbytes = MAX_LFS_FILESIZE;
3971e1a9
AN
230 s->s_blocksize = PAGE_SIZE;
231 s->s_blocksize_bits = PAGE_SHIFT;
1da177e4 232 s->s_magic = magic;
759b9775 233 s->s_op = ops ? ops : &simple_super_operations;
1da177e4
LT
234 s->s_time_gran = 1;
235 root = new_inode(s);
236 if (!root)
237 goto Enomem;
1a1c9bb4
JL
238 /*
239 * since this is the first inode, make it number 1. New inodes created
240 * after this must take care not to collide with it (by passing
241 * max_reserved of 1 to iunique).
242 */
243 root->i_ino = 1;
1da177e4 244 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
1da177e4 245 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
a4464dbc 246 dentry = __d_alloc(s, &d_name);
1da177e4
LT
247 if (!dentry) {
248 iput(root);
249 goto Enomem;
250 }
1da177e4
LT
251 d_instantiate(dentry, root);
252 s->s_root = dentry;
c74a1cbb 253 s->s_d_op = dops;
1da177e4 254 s->s_flags |= MS_ACTIVE;
51139ada 255 return dget(s->s_root);
1da177e4
LT
256
257Enomem:
6f5bbff9 258 deactivate_locked_super(s);
51139ada 259 return ERR_PTR(-ENOMEM);
1da177e4 260}
12f38872 261EXPORT_SYMBOL(mount_pseudo);
1da177e4 262
20955e89
SB
263int simple_open(struct inode *inode, struct file *file)
264{
265 if (inode->i_private)
266 file->private_data = inode->i_private;
267 return 0;
268}
12f38872 269EXPORT_SYMBOL(simple_open);
20955e89 270
1da177e4
LT
271int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
272{
dea655c2 273 struct inode *inode = d_inode(old_dentry);
1da177e4
LT
274
275 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
d8c76e6f 276 inc_nlink(inode);
7de9c6ee 277 ihold(inode);
1da177e4
LT
278 dget(dentry);
279 d_instantiate(dentry, inode);
280 return 0;
281}
12f38872 282EXPORT_SYMBOL(simple_link);
1da177e4 283
1da177e4
LT
284int simple_empty(struct dentry *dentry)
285{
286 struct dentry *child;
287 int ret = 0;
288
2fd6b7f5 289 spin_lock(&dentry->d_lock);
946e51f2 290 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
da502956
NP
291 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
292 if (simple_positive(child)) {
293 spin_unlock(&child->d_lock);
1da177e4 294 goto out;
da502956
NP
295 }
296 spin_unlock(&child->d_lock);
297 }
1da177e4
LT
298 ret = 1;
299out:
2fd6b7f5 300 spin_unlock(&dentry->d_lock);
1da177e4
LT
301 return ret;
302}
12f38872 303EXPORT_SYMBOL(simple_empty);
1da177e4
LT
304
305int simple_unlink(struct inode *dir, struct dentry *dentry)
306{
dea655c2 307 struct inode *inode = d_inode(dentry);
1da177e4
LT
308
309 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
9a53c3a7 310 drop_nlink(inode);
1da177e4
LT
311 dput(dentry);
312 return 0;
313}
12f38872 314EXPORT_SYMBOL(simple_unlink);
1da177e4
LT
315
316int simple_rmdir(struct inode *dir, struct dentry *dentry)
317{
318 if (!simple_empty(dentry))
319 return -ENOTEMPTY;
320
dea655c2 321 drop_nlink(d_inode(dentry));
1da177e4 322 simple_unlink(dir, dentry);
9a53c3a7 323 drop_nlink(dir);
1da177e4
LT
324 return 0;
325}
12f38872 326EXPORT_SYMBOL(simple_rmdir);
1da177e4
LT
327
328int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
329 struct inode *new_dir, struct dentry *new_dentry)
330{
dea655c2 331 struct inode *inode = d_inode(old_dentry);
e36cb0b8 332 int they_are_dirs = d_is_dir(old_dentry);
1da177e4
LT
333
334 if (!simple_empty(new_dentry))
335 return -ENOTEMPTY;
336
dea655c2 337 if (d_really_is_positive(new_dentry)) {
1da177e4 338 simple_unlink(new_dir, new_dentry);
841590ce 339 if (they_are_dirs) {
dea655c2 340 drop_nlink(d_inode(new_dentry));
9a53c3a7 341 drop_nlink(old_dir);
841590ce 342 }
1da177e4 343 } else if (they_are_dirs) {
9a53c3a7 344 drop_nlink(old_dir);
d8c76e6f 345 inc_nlink(new_dir);
1da177e4
LT
346 }
347
348 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
349 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
350
351 return 0;
352}
12f38872 353EXPORT_SYMBOL(simple_rename);
1da177e4 354
7bb46a67 355/**
eef2380c 356 * simple_setattr - setattr for simple filesystem
7bb46a67 357 * @dentry: dentry
358 * @iattr: iattr structure
359 *
360 * Returns 0 on success, -error on failure.
361 *
eef2380c
CH
362 * simple_setattr is a simple ->setattr implementation without a proper
363 * implementation of size changes.
364 *
365 * It can either be used for in-memory filesystems or special files
366 * on simple regular filesystems. Anything that needs to change on-disk
367 * or wire state on size changes needs its own setattr method.
7bb46a67 368 */
369int simple_setattr(struct dentry *dentry, struct iattr *iattr)
370{
dea655c2 371 struct inode *inode = d_inode(dentry);
7bb46a67 372 int error;
373
374 error = inode_change_ok(inode, iattr);
375 if (error)
376 return error;
377
2c27c65e
CH
378 if (iattr->ia_valid & ATTR_SIZE)
379 truncate_setsize(inode, iattr->ia_size);
6a1a90ad 380 setattr_copy(inode, iattr);
eef2380c
CH
381 mark_inode_dirty(inode);
382 return 0;
7bb46a67 383}
384EXPORT_SYMBOL(simple_setattr);
385
1da177e4
LT
386int simple_readpage(struct file *file, struct page *page)
387{
c0d92cbc 388 clear_highpage(page);
1da177e4
LT
389 flush_dcache_page(page);
390 SetPageUptodate(page);
1da177e4
LT
391 unlock_page(page);
392 return 0;
393}
12f38872 394EXPORT_SYMBOL(simple_readpage);
1da177e4 395
afddba49
NP
396int simple_write_begin(struct file *file, struct address_space *mapping,
397 loff_t pos, unsigned len, unsigned flags,
398 struct page **pagep, void **fsdata)
399{
400 struct page *page;
401 pgoff_t index;
afddba49
NP
402
403 index = pos >> PAGE_CACHE_SHIFT;
afddba49 404
54566b2c 405 page = grab_cache_page_write_begin(mapping, index, flags);
afddba49
NP
406 if (!page)
407 return -ENOMEM;
408
409 *pagep = page;
410
193cf4b9
BH
411 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
412 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
413
414 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
415 }
416 return 0;
afddba49 417}
12f38872 418EXPORT_SYMBOL(simple_write_begin);
afddba49 419
ad2a722f
BH
420/**
421 * simple_write_end - .write_end helper for non-block-device FSes
422 * @available: See .write_end of address_space_operations
423 * @file: "
424 * @mapping: "
425 * @pos: "
426 * @len: "
427 * @copied: "
428 * @page: "
429 * @fsdata: "
430 *
431 * simple_write_end does the minimum needed for updating a page after writing is
432 * done. It has the same API signature as the .write_end of
433 * address_space_operations vector. So it can just be set onto .write_end for
434 * FSes that don't need any other processing. i_mutex is assumed to be held.
435 * Block based filesystems should use generic_write_end().
436 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
437 * is not called, so a filesystem that actually does store data in .write_inode
438 * should extend on what's done here with a call to mark_inode_dirty() in the
439 * case that i_size has changed.
440 */
afddba49
NP
441int simple_write_end(struct file *file, struct address_space *mapping,
442 loff_t pos, unsigned len, unsigned copied,
443 struct page *page, void *fsdata)
444{
ad2a722f
BH
445 struct inode *inode = page->mapping->host;
446 loff_t last_pos = pos + copied;
afddba49
NP
447
448 /* zero the stale part of the page if we did a short copy */
449 if (copied < len) {
ad2a722f
BH
450 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
451
452 zero_user(page, from + copied, len - copied);
afddba49
NP
453 }
454
ad2a722f
BH
455 if (!PageUptodate(page))
456 SetPageUptodate(page);
457 /*
458 * No need to use i_size_read() here, the i_size
459 * cannot change under us because we hold the i_mutex.
460 */
461 if (last_pos > inode->i_size)
462 i_size_write(inode, last_pos);
afddba49 463
ad2a722f 464 set_page_dirty(page);
afddba49
NP
465 unlock_page(page);
466 page_cache_release(page);
467
468 return copied;
469}
12f38872 470EXPORT_SYMBOL(simple_write_end);
afddba49 471
1a1c9bb4
JL
472/*
473 * the inodes created here are not hashed. If you use iunique to generate
474 * unique inode values later for this filesystem, then you must take care
475 * to pass it an appropriate max_reserved value to avoid collisions.
476 */
7d683a09
RS
477int simple_fill_super(struct super_block *s, unsigned long magic,
478 struct tree_descr *files)
1da177e4 479{
1da177e4
LT
480 struct inode *inode;
481 struct dentry *root;
482 struct dentry *dentry;
483 int i;
484
485 s->s_blocksize = PAGE_CACHE_SIZE;
486 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
487 s->s_magic = magic;
759b9775 488 s->s_op = &simple_super_operations;
1da177e4
LT
489 s->s_time_gran = 1;
490
491 inode = new_inode(s);
492 if (!inode)
493 return -ENOMEM;
1a1c9bb4
JL
494 /*
495 * because the root inode is 1, the files array must not contain an
496 * entry at index 1
497 */
498 inode->i_ino = 1;
1da177e4 499 inode->i_mode = S_IFDIR | 0755;
1da177e4
LT
500 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
501 inode->i_op = &simple_dir_inode_operations;
502 inode->i_fop = &simple_dir_operations;
bfe86848 503 set_nlink(inode, 2);
48fde701
AV
504 root = d_make_root(inode);
505 if (!root)
1da177e4 506 return -ENOMEM;
1da177e4
LT
507 for (i = 0; !files->name || files->name[0]; i++, files++) {
508 if (!files->name)
509 continue;
1a1c9bb4
JL
510
511 /* warn if it tries to conflict with the root inode */
512 if (unlikely(i == 1))
513 printk(KERN_WARNING "%s: %s passed in a files array"
514 "with an index of 1!\n", __func__,
515 s->s_type->name);
516
1da177e4
LT
517 dentry = d_alloc_name(root, files->name);
518 if (!dentry)
519 goto out;
520 inode = new_inode(s);
32096ea1
KK
521 if (!inode) {
522 dput(dentry);
1da177e4 523 goto out;
32096ea1 524 }
1da177e4 525 inode->i_mode = S_IFREG | files->mode;
1da177e4
LT
526 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
527 inode->i_fop = files->ops;
528 inode->i_ino = i;
529 d_add(dentry, inode);
530 }
531 s->s_root = root;
532 return 0;
533out:
534 d_genocide(root);
640946f2 535 shrink_dcache_parent(root);
1da177e4
LT
536 dput(root);
537 return -ENOMEM;
538}
12f38872 539EXPORT_SYMBOL(simple_fill_super);
1da177e4
LT
540
541static DEFINE_SPINLOCK(pin_fs_lock);
542
1f5ce9e9 543int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
1da177e4
LT
544{
545 struct vfsmount *mnt = NULL;
546 spin_lock(&pin_fs_lock);
547 if (unlikely(!*mount)) {
548 spin_unlock(&pin_fs_lock);
2452992a 549 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
1da177e4
LT
550 if (IS_ERR(mnt))
551 return PTR_ERR(mnt);
552 spin_lock(&pin_fs_lock);
553 if (!*mount)
554 *mount = mnt;
555 }
556 mntget(*mount);
557 ++*count;
558 spin_unlock(&pin_fs_lock);
559 mntput(mnt);
560 return 0;
561}
12f38872 562EXPORT_SYMBOL(simple_pin_fs);
1da177e4
LT
563
564void simple_release_fs(struct vfsmount **mount, int *count)
565{
566 struct vfsmount *mnt;
567 spin_lock(&pin_fs_lock);
568 mnt = *mount;
569 if (!--*count)
570 *mount = NULL;
571 spin_unlock(&pin_fs_lock);
572 mntput(mnt);
573}
12f38872 574EXPORT_SYMBOL(simple_release_fs);
1da177e4 575
6d1029b5
AM
576/**
577 * simple_read_from_buffer - copy data from the buffer to user space
578 * @to: the user space buffer to read to
579 * @count: the maximum number of bytes to read
580 * @ppos: the current position in the buffer
581 * @from: the buffer to read from
582 * @available: the size of the buffer
583 *
584 * The simple_read_from_buffer() function reads up to @count bytes from the
585 * buffer @from at offset @ppos into the user space address starting at @to.
586 *
587 * On success, the number of bytes read is returned and the offset @ppos is
588 * advanced by this number, or negative value is returned on error.
589 **/
1da177e4
LT
590ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
591 const void *from, size_t available)
592{
593 loff_t pos = *ppos;
14be2746
SR
594 size_t ret;
595
1da177e4
LT
596 if (pos < 0)
597 return -EINVAL;
14be2746 598 if (pos >= available || !count)
1da177e4
LT
599 return 0;
600 if (count > available - pos)
601 count = available - pos;
14be2746
SR
602 ret = copy_to_user(to, from + pos, count);
603 if (ret == count)
1da177e4 604 return -EFAULT;
14be2746 605 count -= ret;
1da177e4
LT
606 *ppos = pos + count;
607 return count;
608}
12f38872 609EXPORT_SYMBOL(simple_read_from_buffer);
1da177e4 610
6a727b43
JS
611/**
612 * simple_write_to_buffer - copy data from user space to the buffer
613 * @to: the buffer to write to
614 * @available: the size of the buffer
615 * @ppos: the current position in the buffer
616 * @from: the user space buffer to read from
617 * @count: the maximum number of bytes to read
618 *
619 * The simple_write_to_buffer() function reads up to @count bytes from the user
620 * space address starting at @from into the buffer @to at offset @ppos.
621 *
622 * On success, the number of bytes written is returned and the offset @ppos is
623 * advanced by this number, or negative value is returned on error.
624 **/
625ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
626 const void __user *from, size_t count)
627{
628 loff_t pos = *ppos;
629 size_t res;
630
631 if (pos < 0)
632 return -EINVAL;
633 if (pos >= available || !count)
634 return 0;
635 if (count > available - pos)
636 count = available - pos;
637 res = copy_from_user(to + pos, from, count);
638 if (res == count)
639 return -EFAULT;
640 count -= res;
641 *ppos = pos + count;
642 return count;
643}
12f38872 644EXPORT_SYMBOL(simple_write_to_buffer);
6a727b43 645
6d1029b5
AM
646/**
647 * memory_read_from_buffer - copy data from the buffer
648 * @to: the kernel space buffer to read to
649 * @count: the maximum number of bytes to read
650 * @ppos: the current position in the buffer
651 * @from: the buffer to read from
652 * @available: the size of the buffer
653 *
654 * The memory_read_from_buffer() function reads up to @count bytes from the
655 * buffer @from at offset @ppos into the kernel space address starting at @to.
656 *
657 * On success, the number of bytes read is returned and the offset @ppos is
658 * advanced by this number, or negative value is returned on error.
659 **/
93b07113
AM
660ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
661 const void *from, size_t available)
662{
663 loff_t pos = *ppos;
664
665 if (pos < 0)
666 return -EINVAL;
667 if (pos >= available)
668 return 0;
669 if (count > available - pos)
670 count = available - pos;
671 memcpy(to, from + pos, count);
672 *ppos = pos + count;
673
674 return count;
675}
12f38872 676EXPORT_SYMBOL(memory_read_from_buffer);
93b07113 677
1da177e4
LT
678/*
679 * Transaction based IO.
680 * The file expects a single write which triggers the transaction, and then
681 * possibly a read which collects the result - which is stored in a
682 * file-local buffer.
683 */
76791ab2
IM
684
685void simple_transaction_set(struct file *file, size_t n)
686{
687 struct simple_transaction_argresp *ar = file->private_data;
688
689 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
690
691 /*
692 * The barrier ensures that ar->size will really remain zero until
693 * ar->data is ready for reading.
694 */
695 smp_mb();
696 ar->size = n;
697}
12f38872 698EXPORT_SYMBOL(simple_transaction_set);
76791ab2 699
1da177e4
LT
700char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
701{
702 struct simple_transaction_argresp *ar;
703 static DEFINE_SPINLOCK(simple_transaction_lock);
704
705 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
706 return ERR_PTR(-EFBIG);
707
708 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
709 if (!ar)
710 return ERR_PTR(-ENOMEM);
711
712 spin_lock(&simple_transaction_lock);
713
714 /* only one write allowed per open */
715 if (file->private_data) {
716 spin_unlock(&simple_transaction_lock);
717 free_page((unsigned long)ar);
718 return ERR_PTR(-EBUSY);
719 }
720
721 file->private_data = ar;
722
723 spin_unlock(&simple_transaction_lock);
724
725 if (copy_from_user(ar->data, buf, size))
726 return ERR_PTR(-EFAULT);
727
728 return ar->data;
729}
12f38872 730EXPORT_SYMBOL(simple_transaction_get);
1da177e4
LT
731
732ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
733{
734 struct simple_transaction_argresp *ar = file->private_data;
735
736 if (!ar)
737 return 0;
738 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
739}
12f38872 740EXPORT_SYMBOL(simple_transaction_read);
1da177e4
LT
741
742int simple_transaction_release(struct inode *inode, struct file *file)
743{
744 free_page((unsigned long)file->private_data);
745 return 0;
746}
12f38872 747EXPORT_SYMBOL(simple_transaction_release);
1da177e4 748
acaefc25
AB
749/* Simple attribute files */
750
751struct simple_attr {
8b88b099
CH
752 int (*get)(void *, u64 *);
753 int (*set)(void *, u64);
acaefc25
AB
754 char get_buf[24]; /* enough to store a u64 and "\n\0" */
755 char set_buf[24];
756 void *data;
757 const char *fmt; /* format for read operation */
7cf34c76 758 struct mutex mutex; /* protects access to these buffers */
acaefc25
AB
759};
760
761/* simple_attr_open is called by an actual attribute open file operation
762 * to set the attribute specific access operations. */
763int simple_attr_open(struct inode *inode, struct file *file,
8b88b099 764 int (*get)(void *, u64 *), int (*set)(void *, u64),
acaefc25
AB
765 const char *fmt)
766{
767 struct simple_attr *attr;
768
769 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
770 if (!attr)
771 return -ENOMEM;
772
773 attr->get = get;
774 attr->set = set;
8e18e294 775 attr->data = inode->i_private;
acaefc25 776 attr->fmt = fmt;
7cf34c76 777 mutex_init(&attr->mutex);
acaefc25
AB
778
779 file->private_data = attr;
780
781 return nonseekable_open(inode, file);
782}
12f38872 783EXPORT_SYMBOL_GPL(simple_attr_open);
acaefc25 784
74bedc4d 785int simple_attr_release(struct inode *inode, struct file *file)
acaefc25
AB
786{
787 kfree(file->private_data);
788 return 0;
789}
12f38872 790EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
acaefc25
AB
791
792/* read from the buffer that is filled with the get function */
793ssize_t simple_attr_read(struct file *file, char __user *buf,
794 size_t len, loff_t *ppos)
795{
796 struct simple_attr *attr;
797 size_t size;
798 ssize_t ret;
799
800 attr = file->private_data;
801
802 if (!attr->get)
803 return -EACCES;
804
9261303a
CH
805 ret = mutex_lock_interruptible(&attr->mutex);
806 if (ret)
807 return ret;
808
8b88b099 809 if (*ppos) { /* continued read */
acaefc25 810 size = strlen(attr->get_buf);
8b88b099
CH
811 } else { /* first read */
812 u64 val;
813 ret = attr->get(attr->data, &val);
814 if (ret)
815 goto out;
816
acaefc25 817 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
8b88b099
CH
818 attr->fmt, (unsigned long long)val);
819 }
acaefc25
AB
820
821 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
8b88b099 822out:
7cf34c76 823 mutex_unlock(&attr->mutex);
acaefc25
AB
824 return ret;
825}
12f38872 826EXPORT_SYMBOL_GPL(simple_attr_read);
acaefc25
AB
827
828/* interpret the buffer as a number to call the set function with */
829ssize_t simple_attr_write(struct file *file, const char __user *buf,
830 size_t len, loff_t *ppos)
831{
832 struct simple_attr *attr;
833 u64 val;
834 size_t size;
835 ssize_t ret;
836
837 attr = file->private_data;
acaefc25
AB
838 if (!attr->set)
839 return -EACCES;
840
9261303a
CH
841 ret = mutex_lock_interruptible(&attr->mutex);
842 if (ret)
843 return ret;
844
acaefc25
AB
845 ret = -EFAULT;
846 size = min(sizeof(attr->set_buf) - 1, len);
847 if (copy_from_user(attr->set_buf, buf, size))
848 goto out;
849
acaefc25 850 attr->set_buf[size] = '\0';
f7b88631 851 val = simple_strtoll(attr->set_buf, NULL, 0);
05cc0cee
WF
852 ret = attr->set(attr->data, val);
853 if (ret == 0)
854 ret = len; /* on success, claim we got the whole input */
acaefc25 855out:
7cf34c76 856 mutex_unlock(&attr->mutex);
acaefc25
AB
857 return ret;
858}
12f38872 859EXPORT_SYMBOL_GPL(simple_attr_write);
acaefc25 860
2596110a
CH
861/**
862 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
863 * @sb: filesystem to do the file handle conversion on
864 * @fid: file handle to convert
865 * @fh_len: length of the file handle in bytes
866 * @fh_type: type of file handle
867 * @get_inode: filesystem callback to retrieve inode
868 *
869 * This function decodes @fid as long as it has one of the well-known
870 * Linux filehandle types and calls @get_inode on it to retrieve the
871 * inode for the object specified in the file handle.
872 */
873struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
874 int fh_len, int fh_type, struct inode *(*get_inode)
875 (struct super_block *sb, u64 ino, u32 gen))
876{
877 struct inode *inode = NULL;
878
879 if (fh_len < 2)
880 return NULL;
881
882 switch (fh_type) {
883 case FILEID_INO32_GEN:
884 case FILEID_INO32_GEN_PARENT:
885 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
886 break;
887 }
888
4ea3ada2 889 return d_obtain_alias(inode);
2596110a
CH
890}
891EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
892
893/**
ca186830 894 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
2596110a
CH
895 * @sb: filesystem to do the file handle conversion on
896 * @fid: file handle to convert
897 * @fh_len: length of the file handle in bytes
898 * @fh_type: type of file handle
899 * @get_inode: filesystem callback to retrieve inode
900 *
901 * This function decodes @fid as long as it has one of the well-known
902 * Linux filehandle types and calls @get_inode on it to retrieve the
903 * inode for the _parent_ object specified in the file handle if it
904 * is specified in the file handle, or NULL otherwise.
905 */
906struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
907 int fh_len, int fh_type, struct inode *(*get_inode)
908 (struct super_block *sb, u64 ino, u32 gen))
909{
910 struct inode *inode = NULL;
911
912 if (fh_len <= 2)
913 return NULL;
914
915 switch (fh_type) {
916 case FILEID_INO32_GEN_PARENT:
917 inode = get_inode(sb, fid->i32.parent_ino,
918 (fh_len > 3 ? fid->i32.parent_gen : 0));
919 break;
920 }
921
4ea3ada2 922 return d_obtain_alias(inode);
2596110a
CH
923}
924EXPORT_SYMBOL_GPL(generic_fh_to_parent);
925
1b061d92 926/**
ac13a829
FF
927 * __generic_file_fsync - generic fsync implementation for simple filesystems
928 *
1b061d92 929 * @file: file to synchronize
ac13a829
FF
930 * @start: start offset in bytes
931 * @end: end offset in bytes (inclusive)
1b061d92
CH
932 * @datasync: only synchronize essential metadata if true
933 *
934 * This is a generic implementation of the fsync method for simple
935 * filesystems which track all non-inode metadata in the buffers list
936 * hanging off the address_space structure.
937 */
ac13a829
FF
938int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
939 int datasync)
d5aacad5 940{
7ea80859 941 struct inode *inode = file->f_mapping->host;
d5aacad5
AV
942 int err;
943 int ret;
944
02c24a82
JB
945 err = filemap_write_and_wait_range(inode->i_mapping, start, end);
946 if (err)
947 return err;
948
949 mutex_lock(&inode->i_mutex);
d5aacad5 950 ret = sync_mapping_buffers(inode->i_mapping);
0ae45f63 951 if (!(inode->i_state & I_DIRTY_ALL))
02c24a82 952 goto out;
d5aacad5 953 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
02c24a82 954 goto out;
d5aacad5 955
c3765016 956 err = sync_inode_metadata(inode, 1);
d5aacad5
AV
957 if (ret == 0)
958 ret = err;
ac13a829 959
02c24a82
JB
960out:
961 mutex_unlock(&inode->i_mutex);
d5aacad5
AV
962 return ret;
963}
ac13a829
FF
964EXPORT_SYMBOL(__generic_file_fsync);
965
966/**
967 * generic_file_fsync - generic fsync implementation for simple filesystems
968 * with flush
969 * @file: file to synchronize
970 * @start: start offset in bytes
971 * @end: end offset in bytes (inclusive)
972 * @datasync: only synchronize essential metadata if true
973 *
974 */
975
976int generic_file_fsync(struct file *file, loff_t start, loff_t end,
977 int datasync)
978{
979 struct inode *inode = file->f_mapping->host;
980 int err;
981
982 err = __generic_file_fsync(file, start, end, datasync);
983 if (err)
984 return err;
985 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
986}
1b061d92
CH
987EXPORT_SYMBOL(generic_file_fsync);
988
30ca22c7
PL
989/**
990 * generic_check_addressable - Check addressability of file system
991 * @blocksize_bits: log of file system block size
992 * @num_blocks: number of blocks in file system
993 *
994 * Determine whether a file system with @num_blocks blocks (and a
995 * block size of 2**@blocksize_bits) is addressable by the sector_t
996 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
997 */
998int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
999{
1000 u64 last_fs_block = num_blocks - 1;
a33f13ef
JB
1001 u64 last_fs_page =
1002 last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
30ca22c7
PL
1003
1004 if (unlikely(num_blocks == 0))
1005 return 0;
1006
1007 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
1008 return -EINVAL;
1009
a33f13ef
JB
1010 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1011 (last_fs_page > (pgoff_t)(~0ULL))) {
30ca22c7
PL
1012 return -EFBIG;
1013 }
1014 return 0;
1015}
1016EXPORT_SYMBOL(generic_check_addressable);
1017
1b061d92
CH
1018/*
1019 * No-op implementation of ->fsync for in-memory filesystems.
1020 */
02c24a82 1021int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1b061d92
CH
1022{
1023 return 0;
1024}
1b061d92 1025EXPORT_SYMBOL(noop_fsync);
87dc800b 1026
680baacb 1027void kfree_put_link(struct dentry *dentry, void *cookie)
87dc800b 1028{
680baacb 1029 kfree(cookie);
87dc800b
AV
1030}
1031EXPORT_SYMBOL(kfree_put_link);
6987843f
AV
1032
1033/*
1034 * nop .set_page_dirty method so that people can use .page_mkwrite on
1035 * anon inodes.
1036 */
1037static int anon_set_page_dirty(struct page *page)
1038{
1039 return 0;
1040};
1041
1042/*
1043 * A single inode exists for all anon_inode files. Contrary to pipes,
1044 * anon_inode inodes have no associated per-instance data, so we need
1045 * only allocate one of them.
1046 */
1047struct inode *alloc_anon_inode(struct super_block *s)
1048{
1049 static const struct address_space_operations anon_aops = {
1050 .set_page_dirty = anon_set_page_dirty,
1051 };
1052 struct inode *inode = new_inode_pseudo(s);
1053
1054 if (!inode)
1055 return ERR_PTR(-ENOMEM);
1056
1057 inode->i_ino = get_next_ino();
1058 inode->i_mapping->a_ops = &anon_aops;
1059
1060 /*
1061 * Mark the inode dirty from the very beginning,
1062 * that way it will never be moved to the dirty
1063 * list because mark_inode_dirty() will think
1064 * that it already _is_ on the dirty list.
1065 */
1066 inode->i_state = I_DIRTY;
1067 inode->i_mode = S_IRUSR | S_IWUSR;
1068 inode->i_uid = current_fsuid();
1069 inode->i_gid = current_fsgid();
1070 inode->i_flags |= S_PRIVATE;
1071 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1072 return inode;
1073}
1074EXPORT_SYMBOL(alloc_anon_inode);
1c994a09
JL
1075
1076/**
1077 * simple_nosetlease - generic helper for prohibiting leases
1078 * @filp: file pointer
1079 * @arg: type of lease to obtain
1080 * @flp: new lease supplied for insertion
e6f5c789 1081 * @priv: private data for lm_setup operation
1c994a09
JL
1082 *
1083 * Generic helper for filesystems that do not wish to allow leases to be set.
1084 * All arguments are ignored and it just returns -EINVAL.
1085 */
1086int
e6f5c789
JL
1087simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1088 void **priv)
1c994a09
JL
1089{
1090 return -EINVAL;
1091}
1092EXPORT_SYMBOL(simple_nosetlease);
61ba64fc 1093
680baacb 1094const char *simple_follow_link(struct dentry *dentry, void **cookie, struct nameidata *nd)
61ba64fc 1095{
680baacb 1096 return d_inode(dentry)->i_link;
61ba64fc
AV
1097}
1098EXPORT_SYMBOL(simple_follow_link);
1099
1100const struct inode_operations simple_symlink_inode_operations = {
1101 .follow_link = simple_follow_link,
1102 .readlink = generic_readlink
1103};
1104EXPORT_SYMBOL(simple_symlink_inode_operations);