]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/libfs.c
OCFS2: Allow huge (> 16 TiB) volumes to mount
[net-next-2.6.git] / fs / libfs.c
CommitLineData
1da177e4
LT
1/*
2 * fs/libfs.c
3 * Library for filesystems writers.
4 */
5
6#include <linux/module.h>
7#include <linux/pagemap.h>
5a0e3ad6 8#include <linux/slab.h>
1da177e4
LT
9#include <linux/mount.h>
10#include <linux/vfs.h>
7bb46a67 11#include <linux/quotaops.h>
7cf34c76 12#include <linux/mutex.h>
2596110a 13#include <linux/exportfs.h>
d5aacad5
AV
14#include <linux/writeback.h>
15#include <linux/buffer_head.h>
7cf34c76 16
1da177e4
LT
17#include <asm/uaccess.h>
18
19int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
20 struct kstat *stat)
21{
22 struct inode *inode = dentry->d_inode;
23 generic_fillattr(inode, stat);
24 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
25 return 0;
26}
27
726c3342 28int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
1da177e4 29{
726c3342 30 buf->f_type = dentry->d_sb->s_magic;
1da177e4
LT
31 buf->f_bsize = PAGE_CACHE_SIZE;
32 buf->f_namelen = NAME_MAX;
33 return 0;
34}
35
36/*
37 * Retaining negative dentries for an in-memory filesystem just wastes
38 * memory and lookup time: arrange for them to be deleted immediately.
39 */
40static int simple_delete_dentry(struct dentry *dentry)
41{
42 return 1;
43}
44
45/*
46 * Lookup the data. This is trivial - if the dentry didn't already
47 * exist, we know it is negative. Set d_op to delete negative dentries.
48 */
49struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
50{
3ba13d17 51 static const struct dentry_operations simple_dentry_operations = {
1da177e4
LT
52 .d_delete = simple_delete_dentry,
53 };
54
55 if (dentry->d_name.len > NAME_MAX)
56 return ERR_PTR(-ENAMETOOLONG);
57 dentry->d_op = &simple_dentry_operations;
58 d_add(dentry, NULL);
59 return NULL;
60}
61
1da177e4
LT
62int dcache_dir_open(struct inode *inode, struct file *file)
63{
64 static struct qstr cursor_name = {.len = 1, .name = "."};
65
0f7fc9e4 66 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
1da177e4
LT
67
68 return file->private_data ? 0 : -ENOMEM;
69}
70
71int dcache_dir_close(struct inode *inode, struct file *file)
72{
73 dput(file->private_data);
74 return 0;
75}
76
77loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
78{
0f7fc9e4 79 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
1da177e4
LT
80 switch (origin) {
81 case 1:
82 offset += file->f_pos;
83 case 0:
84 if (offset >= 0)
85 break;
86 default:
0f7fc9e4 87 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
1da177e4
LT
88 return -EINVAL;
89 }
90 if (offset != file->f_pos) {
91 file->f_pos = offset;
92 if (file->f_pos >= 2) {
93 struct list_head *p;
94 struct dentry *cursor = file->private_data;
95 loff_t n = file->f_pos - 2;
96
97 spin_lock(&dcache_lock);
5160ee6f 98 list_del(&cursor->d_u.d_child);
0f7fc9e4
JJS
99 p = file->f_path.dentry->d_subdirs.next;
100 while (n && p != &file->f_path.dentry->d_subdirs) {
1da177e4 101 struct dentry *next;
5160ee6f 102 next = list_entry(p, struct dentry, d_u.d_child);
1da177e4
LT
103 if (!d_unhashed(next) && next->d_inode)
104 n--;
105 p = p->next;
106 }
5160ee6f 107 list_add_tail(&cursor->d_u.d_child, p);
1da177e4
LT
108 spin_unlock(&dcache_lock);
109 }
110 }
0f7fc9e4 111 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
1da177e4
LT
112 return offset;
113}
114
115/* Relationship between i_mode and the DT_xxx types */
116static inline unsigned char dt_type(struct inode *inode)
117{
118 return (inode->i_mode >> 12) & 15;
119}
120
121/*
122 * Directory is locked and all positive dentries in it are safe, since
123 * for ramfs-type trees they can't go away without unlink() or rmdir(),
124 * both impossible due to the lock on directory.
125 */
126
127int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
128{
0f7fc9e4 129 struct dentry *dentry = filp->f_path.dentry;
1da177e4 130 struct dentry *cursor = filp->private_data;
5160ee6f 131 struct list_head *p, *q = &cursor->d_u.d_child;
1da177e4
LT
132 ino_t ino;
133 int i = filp->f_pos;
134
135 switch (i) {
136 case 0:
137 ino = dentry->d_inode->i_ino;
138 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
139 break;
140 filp->f_pos++;
141 i++;
142 /* fallthrough */
143 case 1:
144 ino = parent_ino(dentry);
145 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
146 break;
147 filp->f_pos++;
148 i++;
149 /* fallthrough */
150 default:
151 spin_lock(&dcache_lock);
1bfba4e8
AM
152 if (filp->f_pos == 2)
153 list_move(q, &dentry->d_subdirs);
154
1da177e4
LT
155 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
156 struct dentry *next;
5160ee6f 157 next = list_entry(p, struct dentry, d_u.d_child);
1da177e4
LT
158 if (d_unhashed(next) || !next->d_inode)
159 continue;
160
161 spin_unlock(&dcache_lock);
0f8952c2
RN
162 if (filldir(dirent, next->d_name.name,
163 next->d_name.len, filp->f_pos,
164 next->d_inode->i_ino,
165 dt_type(next->d_inode)) < 0)
1da177e4
LT
166 return 0;
167 spin_lock(&dcache_lock);
168 /* next is still alive */
1bfba4e8 169 list_move(q, p);
1da177e4
LT
170 p = q;
171 filp->f_pos++;
172 }
173 spin_unlock(&dcache_lock);
174 }
175 return 0;
176}
177
178ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
179{
180 return -EISDIR;
181}
182
4b6f5d20 183const struct file_operations simple_dir_operations = {
1da177e4
LT
184 .open = dcache_dir_open,
185 .release = dcache_dir_close,
186 .llseek = dcache_dir_lseek,
187 .read = generic_read_dir,
188 .readdir = dcache_readdir,
1b061d92 189 .fsync = noop_fsync,
1da177e4
LT
190};
191
92e1d5be 192const struct inode_operations simple_dir_inode_operations = {
1da177e4
LT
193 .lookup = simple_lookup,
194};
195
759b9775
HD
196static const struct super_operations simple_super_operations = {
197 .statfs = simple_statfs,
198};
199
1da177e4
LT
200/*
201 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
202 * will never be mountable)
203 */
454e2398 204int get_sb_pseudo(struct file_system_type *fs_type, char *name,
ee9b6d61 205 const struct super_operations *ops, unsigned long magic,
454e2398 206 struct vfsmount *mnt)
1da177e4
LT
207{
208 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
1da177e4
LT
209 struct dentry *dentry;
210 struct inode *root;
211 struct qstr d_name = {.name = name, .len = strlen(name)};
212
213 if (IS_ERR(s))
454e2398 214 return PTR_ERR(s);
1da177e4
LT
215
216 s->s_flags = MS_NOUSER;
89a4eb4b 217 s->s_maxbytes = MAX_LFS_FILESIZE;
3971e1a9
AN
218 s->s_blocksize = PAGE_SIZE;
219 s->s_blocksize_bits = PAGE_SHIFT;
1da177e4 220 s->s_magic = magic;
759b9775 221 s->s_op = ops ? ops : &simple_super_operations;
1da177e4
LT
222 s->s_time_gran = 1;
223 root = new_inode(s);
224 if (!root)
225 goto Enomem;
1a1c9bb4
JL
226 /*
227 * since this is the first inode, make it number 1. New inodes created
228 * after this must take care not to collide with it (by passing
229 * max_reserved of 1 to iunique).
230 */
231 root->i_ino = 1;
1da177e4 232 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
1da177e4
LT
233 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
234 dentry = d_alloc(NULL, &d_name);
235 if (!dentry) {
236 iput(root);
237 goto Enomem;
238 }
239 dentry->d_sb = s;
240 dentry->d_parent = dentry;
241 d_instantiate(dentry, root);
242 s->s_root = dentry;
243 s->s_flags |= MS_ACTIVE;
a3ec947c
SB
244 simple_set_mnt(mnt, s);
245 return 0;
1da177e4
LT
246
247Enomem:
6f5bbff9 248 deactivate_locked_super(s);
454e2398 249 return -ENOMEM;
1da177e4
LT
250}
251
252int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
253{
254 struct inode *inode = old_dentry->d_inode;
255
256 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
d8c76e6f 257 inc_nlink(inode);
1da177e4
LT
258 atomic_inc(&inode->i_count);
259 dget(dentry);
260 d_instantiate(dentry, inode);
261 return 0;
262}
263
264static inline int simple_positive(struct dentry *dentry)
265{
266 return dentry->d_inode && !d_unhashed(dentry);
267}
268
269int simple_empty(struct dentry *dentry)
270{
271 struct dentry *child;
272 int ret = 0;
273
274 spin_lock(&dcache_lock);
5160ee6f 275 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
1da177e4
LT
276 if (simple_positive(child))
277 goto out;
278 ret = 1;
279out:
280 spin_unlock(&dcache_lock);
281 return ret;
282}
283
284int simple_unlink(struct inode *dir, struct dentry *dentry)
285{
286 struct inode *inode = dentry->d_inode;
287
288 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
9a53c3a7 289 drop_nlink(inode);
1da177e4
LT
290 dput(dentry);
291 return 0;
292}
293
294int simple_rmdir(struct inode *dir, struct dentry *dentry)
295{
296 if (!simple_empty(dentry))
297 return -ENOTEMPTY;
298
9a53c3a7 299 drop_nlink(dentry->d_inode);
1da177e4 300 simple_unlink(dir, dentry);
9a53c3a7 301 drop_nlink(dir);
1da177e4
LT
302 return 0;
303}
304
305int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
306 struct inode *new_dir, struct dentry *new_dentry)
307{
308 struct inode *inode = old_dentry->d_inode;
309 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
310
311 if (!simple_empty(new_dentry))
312 return -ENOTEMPTY;
313
314 if (new_dentry->d_inode) {
315 simple_unlink(new_dir, new_dentry);
316 if (they_are_dirs)
9a53c3a7 317 drop_nlink(old_dir);
1da177e4 318 } else if (they_are_dirs) {
9a53c3a7 319 drop_nlink(old_dir);
d8c76e6f 320 inc_nlink(new_dir);
1da177e4
LT
321 }
322
323 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
324 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
325
326 return 0;
327}
328
7bb46a67 329/**
eef2380c 330 * simple_setattr - setattr for simple filesystem
7bb46a67 331 * @dentry: dentry
332 * @iattr: iattr structure
333 *
334 * Returns 0 on success, -error on failure.
335 *
eef2380c
CH
336 * simple_setattr is a simple ->setattr implementation without a proper
337 * implementation of size changes.
338 *
339 * It can either be used for in-memory filesystems or special files
340 * on simple regular filesystems. Anything that needs to change on-disk
341 * or wire state on size changes needs its own setattr method.
7bb46a67 342 */
343int simple_setattr(struct dentry *dentry, struct iattr *iattr)
344{
345 struct inode *inode = dentry->d_inode;
346 int error;
347
eef2380c
CH
348 WARN_ON_ONCE(inode->i_op->truncate);
349
7bb46a67 350 error = inode_change_ok(inode, iattr);
351 if (error)
352 return error;
353
2c27c65e
CH
354 if (iattr->ia_valid & ATTR_SIZE)
355 truncate_setsize(inode, iattr->ia_size);
6a1a90ad 356 setattr_copy(inode, iattr);
eef2380c
CH
357 mark_inode_dirty(inode);
358 return 0;
7bb46a67 359}
360EXPORT_SYMBOL(simple_setattr);
361
1da177e4
LT
362int simple_readpage(struct file *file, struct page *page)
363{
c0d92cbc 364 clear_highpage(page);
1da177e4
LT
365 flush_dcache_page(page);
366 SetPageUptodate(page);
1da177e4
LT
367 unlock_page(page);
368 return 0;
369}
370
afddba49
NP
371int simple_write_begin(struct file *file, struct address_space *mapping,
372 loff_t pos, unsigned len, unsigned flags,
373 struct page **pagep, void **fsdata)
374{
375 struct page *page;
376 pgoff_t index;
afddba49
NP
377
378 index = pos >> PAGE_CACHE_SHIFT;
afddba49 379
54566b2c 380 page = grab_cache_page_write_begin(mapping, index, flags);
afddba49
NP
381 if (!page)
382 return -ENOMEM;
383
384 *pagep = page;
385
193cf4b9
BH
386 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
387 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
388
389 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
390 }
391 return 0;
afddba49
NP
392}
393
ad2a722f
BH
394/**
395 * simple_write_end - .write_end helper for non-block-device FSes
396 * @available: See .write_end of address_space_operations
397 * @file: "
398 * @mapping: "
399 * @pos: "
400 * @len: "
401 * @copied: "
402 * @page: "
403 * @fsdata: "
404 *
405 * simple_write_end does the minimum needed for updating a page after writing is
406 * done. It has the same API signature as the .write_end of
407 * address_space_operations vector. So it can just be set onto .write_end for
408 * FSes that don't need any other processing. i_mutex is assumed to be held.
409 * Block based filesystems should use generic_write_end().
410 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
411 * is not called, so a filesystem that actually does store data in .write_inode
412 * should extend on what's done here with a call to mark_inode_dirty() in the
413 * case that i_size has changed.
414 */
afddba49
NP
415int simple_write_end(struct file *file, struct address_space *mapping,
416 loff_t pos, unsigned len, unsigned copied,
417 struct page *page, void *fsdata)
418{
ad2a722f
BH
419 struct inode *inode = page->mapping->host;
420 loff_t last_pos = pos + copied;
afddba49
NP
421
422 /* zero the stale part of the page if we did a short copy */
423 if (copied < len) {
ad2a722f
BH
424 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
425
426 zero_user(page, from + copied, len - copied);
afddba49
NP
427 }
428
ad2a722f
BH
429 if (!PageUptodate(page))
430 SetPageUptodate(page);
431 /*
432 * No need to use i_size_read() here, the i_size
433 * cannot change under us because we hold the i_mutex.
434 */
435 if (last_pos > inode->i_size)
436 i_size_write(inode, last_pos);
afddba49 437
ad2a722f 438 set_page_dirty(page);
afddba49
NP
439 unlock_page(page);
440 page_cache_release(page);
441
442 return copied;
443}
444
1a1c9bb4
JL
445/*
446 * the inodes created here are not hashed. If you use iunique to generate
447 * unique inode values later for this filesystem, then you must take care
448 * to pass it an appropriate max_reserved value to avoid collisions.
449 */
7d683a09
RS
450int simple_fill_super(struct super_block *s, unsigned long magic,
451 struct tree_descr *files)
1da177e4 452{
1da177e4
LT
453 struct inode *inode;
454 struct dentry *root;
455 struct dentry *dentry;
456 int i;
457
458 s->s_blocksize = PAGE_CACHE_SIZE;
459 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
460 s->s_magic = magic;
759b9775 461 s->s_op = &simple_super_operations;
1da177e4
LT
462 s->s_time_gran = 1;
463
464 inode = new_inode(s);
465 if (!inode)
466 return -ENOMEM;
1a1c9bb4
JL
467 /*
468 * because the root inode is 1, the files array must not contain an
469 * entry at index 1
470 */
471 inode->i_ino = 1;
1da177e4 472 inode->i_mode = S_IFDIR | 0755;
1da177e4
LT
473 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
474 inode->i_op = &simple_dir_inode_operations;
475 inode->i_fop = &simple_dir_operations;
7656f328 476 inode->i_nlink = 2;
1da177e4
LT
477 root = d_alloc_root(inode);
478 if (!root) {
479 iput(inode);
480 return -ENOMEM;
481 }
482 for (i = 0; !files->name || files->name[0]; i++, files++) {
483 if (!files->name)
484 continue;
1a1c9bb4
JL
485
486 /* warn if it tries to conflict with the root inode */
487 if (unlikely(i == 1))
488 printk(KERN_WARNING "%s: %s passed in a files array"
489 "with an index of 1!\n", __func__,
490 s->s_type->name);
491
1da177e4
LT
492 dentry = d_alloc_name(root, files->name);
493 if (!dentry)
494 goto out;
495 inode = new_inode(s);
496 if (!inode)
497 goto out;
498 inode->i_mode = S_IFREG | files->mode;
1da177e4
LT
499 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
500 inode->i_fop = files->ops;
501 inode->i_ino = i;
502 d_add(dentry, inode);
503 }
504 s->s_root = root;
505 return 0;
506out:
507 d_genocide(root);
508 dput(root);
509 return -ENOMEM;
510}
511
512static DEFINE_SPINLOCK(pin_fs_lock);
513
1f5ce9e9 514int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
1da177e4
LT
515{
516 struct vfsmount *mnt = NULL;
517 spin_lock(&pin_fs_lock);
518 if (unlikely(!*mount)) {
519 spin_unlock(&pin_fs_lock);
1f5ce9e9 520 mnt = vfs_kern_mount(type, 0, type->name, NULL);
1da177e4
LT
521 if (IS_ERR(mnt))
522 return PTR_ERR(mnt);
523 spin_lock(&pin_fs_lock);
524 if (!*mount)
525 *mount = mnt;
526 }
527 mntget(*mount);
528 ++*count;
529 spin_unlock(&pin_fs_lock);
530 mntput(mnt);
531 return 0;
532}
533
534void simple_release_fs(struct vfsmount **mount, int *count)
535{
536 struct vfsmount *mnt;
537 spin_lock(&pin_fs_lock);
538 mnt = *mount;
539 if (!--*count)
540 *mount = NULL;
541 spin_unlock(&pin_fs_lock);
542 mntput(mnt);
543}
544
6d1029b5
AM
545/**
546 * simple_read_from_buffer - copy data from the buffer to user space
547 * @to: the user space buffer to read to
548 * @count: the maximum number of bytes to read
549 * @ppos: the current position in the buffer
550 * @from: the buffer to read from
551 * @available: the size of the buffer
552 *
553 * The simple_read_from_buffer() function reads up to @count bytes from the
554 * buffer @from at offset @ppos into the user space address starting at @to.
555 *
556 * On success, the number of bytes read is returned and the offset @ppos is
557 * advanced by this number, or negative value is returned on error.
558 **/
1da177e4
LT
559ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
560 const void *from, size_t available)
561{
562 loff_t pos = *ppos;
14be2746
SR
563 size_t ret;
564
1da177e4
LT
565 if (pos < 0)
566 return -EINVAL;
14be2746 567 if (pos >= available || !count)
1da177e4
LT
568 return 0;
569 if (count > available - pos)
570 count = available - pos;
14be2746
SR
571 ret = copy_to_user(to, from + pos, count);
572 if (ret == count)
1da177e4 573 return -EFAULT;
14be2746 574 count -= ret;
1da177e4
LT
575 *ppos = pos + count;
576 return count;
577}
578
6a727b43
JS
579/**
580 * simple_write_to_buffer - copy data from user space to the buffer
581 * @to: the buffer to write to
582 * @available: the size of the buffer
583 * @ppos: the current position in the buffer
584 * @from: the user space buffer to read from
585 * @count: the maximum number of bytes to read
586 *
587 * The simple_write_to_buffer() function reads up to @count bytes from the user
588 * space address starting at @from into the buffer @to at offset @ppos.
589 *
590 * On success, the number of bytes written is returned and the offset @ppos is
591 * advanced by this number, or negative value is returned on error.
592 **/
593ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
594 const void __user *from, size_t count)
595{
596 loff_t pos = *ppos;
597 size_t res;
598
599 if (pos < 0)
600 return -EINVAL;
601 if (pos >= available || !count)
602 return 0;
603 if (count > available - pos)
604 count = available - pos;
605 res = copy_from_user(to + pos, from, count);
606 if (res == count)
607 return -EFAULT;
608 count -= res;
609 *ppos = pos + count;
610 return count;
611}
612
6d1029b5
AM
613/**
614 * memory_read_from_buffer - copy data from the buffer
615 * @to: the kernel space buffer to read to
616 * @count: the maximum number of bytes to read
617 * @ppos: the current position in the buffer
618 * @from: the buffer to read from
619 * @available: the size of the buffer
620 *
621 * The memory_read_from_buffer() function reads up to @count bytes from the
622 * buffer @from at offset @ppos into the kernel space address starting at @to.
623 *
624 * On success, the number of bytes read is returned and the offset @ppos is
625 * advanced by this number, or negative value is returned on error.
626 **/
93b07113
AM
627ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
628 const void *from, size_t available)
629{
630 loff_t pos = *ppos;
631
632 if (pos < 0)
633 return -EINVAL;
634 if (pos >= available)
635 return 0;
636 if (count > available - pos)
637 count = available - pos;
638 memcpy(to, from + pos, count);
639 *ppos = pos + count;
640
641 return count;
642}
643
1da177e4
LT
644/*
645 * Transaction based IO.
646 * The file expects a single write which triggers the transaction, and then
647 * possibly a read which collects the result - which is stored in a
648 * file-local buffer.
649 */
76791ab2
IM
650
651void simple_transaction_set(struct file *file, size_t n)
652{
653 struct simple_transaction_argresp *ar = file->private_data;
654
655 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
656
657 /*
658 * The barrier ensures that ar->size will really remain zero until
659 * ar->data is ready for reading.
660 */
661 smp_mb();
662 ar->size = n;
663}
664
1da177e4
LT
665char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
666{
667 struct simple_transaction_argresp *ar;
668 static DEFINE_SPINLOCK(simple_transaction_lock);
669
670 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
671 return ERR_PTR(-EFBIG);
672
673 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
674 if (!ar)
675 return ERR_PTR(-ENOMEM);
676
677 spin_lock(&simple_transaction_lock);
678
679 /* only one write allowed per open */
680 if (file->private_data) {
681 spin_unlock(&simple_transaction_lock);
682 free_page((unsigned long)ar);
683 return ERR_PTR(-EBUSY);
684 }
685
686 file->private_data = ar;
687
688 spin_unlock(&simple_transaction_lock);
689
690 if (copy_from_user(ar->data, buf, size))
691 return ERR_PTR(-EFAULT);
692
693 return ar->data;
694}
695
696ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
697{
698 struct simple_transaction_argresp *ar = file->private_data;
699
700 if (!ar)
701 return 0;
702 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
703}
704
705int simple_transaction_release(struct inode *inode, struct file *file)
706{
707 free_page((unsigned long)file->private_data);
708 return 0;
709}
710
acaefc25
AB
711/* Simple attribute files */
712
713struct simple_attr {
8b88b099
CH
714 int (*get)(void *, u64 *);
715 int (*set)(void *, u64);
acaefc25
AB
716 char get_buf[24]; /* enough to store a u64 and "\n\0" */
717 char set_buf[24];
718 void *data;
719 const char *fmt; /* format for read operation */
7cf34c76 720 struct mutex mutex; /* protects access to these buffers */
acaefc25
AB
721};
722
723/* simple_attr_open is called by an actual attribute open file operation
724 * to set the attribute specific access operations. */
725int simple_attr_open(struct inode *inode, struct file *file,
8b88b099 726 int (*get)(void *, u64 *), int (*set)(void *, u64),
acaefc25
AB
727 const char *fmt)
728{
729 struct simple_attr *attr;
730
731 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
732 if (!attr)
733 return -ENOMEM;
734
735 attr->get = get;
736 attr->set = set;
8e18e294 737 attr->data = inode->i_private;
acaefc25 738 attr->fmt = fmt;
7cf34c76 739 mutex_init(&attr->mutex);
acaefc25
AB
740
741 file->private_data = attr;
742
743 return nonseekable_open(inode, file);
744}
745
74bedc4d 746int simple_attr_release(struct inode *inode, struct file *file)
acaefc25
AB
747{
748 kfree(file->private_data);
749 return 0;
750}
751
752/* read from the buffer that is filled with the get function */
753ssize_t simple_attr_read(struct file *file, char __user *buf,
754 size_t len, loff_t *ppos)
755{
756 struct simple_attr *attr;
757 size_t size;
758 ssize_t ret;
759
760 attr = file->private_data;
761
762 if (!attr->get)
763 return -EACCES;
764
9261303a
CH
765 ret = mutex_lock_interruptible(&attr->mutex);
766 if (ret)
767 return ret;
768
8b88b099 769 if (*ppos) { /* continued read */
acaefc25 770 size = strlen(attr->get_buf);
8b88b099
CH
771 } else { /* first read */
772 u64 val;
773 ret = attr->get(attr->data, &val);
774 if (ret)
775 goto out;
776
acaefc25 777 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
8b88b099
CH
778 attr->fmt, (unsigned long long)val);
779 }
acaefc25
AB
780
781 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
8b88b099 782out:
7cf34c76 783 mutex_unlock(&attr->mutex);
acaefc25
AB
784 return ret;
785}
786
787/* interpret the buffer as a number to call the set function with */
788ssize_t simple_attr_write(struct file *file, const char __user *buf,
789 size_t len, loff_t *ppos)
790{
791 struct simple_attr *attr;
792 u64 val;
793 size_t size;
794 ssize_t ret;
795
796 attr = file->private_data;
acaefc25
AB
797 if (!attr->set)
798 return -EACCES;
799
9261303a
CH
800 ret = mutex_lock_interruptible(&attr->mutex);
801 if (ret)
802 return ret;
803
acaefc25
AB
804 ret = -EFAULT;
805 size = min(sizeof(attr->set_buf) - 1, len);
806 if (copy_from_user(attr->set_buf, buf, size))
807 goto out;
808
acaefc25
AB
809 attr->set_buf[size] = '\0';
810 val = simple_strtol(attr->set_buf, NULL, 0);
05cc0cee
WF
811 ret = attr->set(attr->data, val);
812 if (ret == 0)
813 ret = len; /* on success, claim we got the whole input */
acaefc25 814out:
7cf34c76 815 mutex_unlock(&attr->mutex);
acaefc25
AB
816 return ret;
817}
818
2596110a
CH
819/**
820 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
821 * @sb: filesystem to do the file handle conversion on
822 * @fid: file handle to convert
823 * @fh_len: length of the file handle in bytes
824 * @fh_type: type of file handle
825 * @get_inode: filesystem callback to retrieve inode
826 *
827 * This function decodes @fid as long as it has one of the well-known
828 * Linux filehandle types and calls @get_inode on it to retrieve the
829 * inode for the object specified in the file handle.
830 */
831struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
832 int fh_len, int fh_type, struct inode *(*get_inode)
833 (struct super_block *sb, u64 ino, u32 gen))
834{
835 struct inode *inode = NULL;
836
837 if (fh_len < 2)
838 return NULL;
839
840 switch (fh_type) {
841 case FILEID_INO32_GEN:
842 case FILEID_INO32_GEN_PARENT:
843 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
844 break;
845 }
846
4ea3ada2 847 return d_obtain_alias(inode);
2596110a
CH
848}
849EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
850
851/**
852 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
853 * @sb: filesystem to do the file handle conversion on
854 * @fid: file handle to convert
855 * @fh_len: length of the file handle in bytes
856 * @fh_type: type of file handle
857 * @get_inode: filesystem callback to retrieve inode
858 *
859 * This function decodes @fid as long as it has one of the well-known
860 * Linux filehandle types and calls @get_inode on it to retrieve the
861 * inode for the _parent_ object specified in the file handle if it
862 * is specified in the file handle, or NULL otherwise.
863 */
864struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
865 int fh_len, int fh_type, struct inode *(*get_inode)
866 (struct super_block *sb, u64 ino, u32 gen))
867{
868 struct inode *inode = NULL;
869
870 if (fh_len <= 2)
871 return NULL;
872
873 switch (fh_type) {
874 case FILEID_INO32_GEN_PARENT:
875 inode = get_inode(sb, fid->i32.parent_ino,
876 (fh_len > 3 ? fid->i32.parent_gen : 0));
877 break;
878 }
879
4ea3ada2 880 return d_obtain_alias(inode);
2596110a
CH
881}
882EXPORT_SYMBOL_GPL(generic_fh_to_parent);
883
1b061d92
CH
884/**
885 * generic_file_fsync - generic fsync implementation for simple filesystems
886 * @file: file to synchronize
887 * @datasync: only synchronize essential metadata if true
888 *
889 * This is a generic implementation of the fsync method for simple
890 * filesystems which track all non-inode metadata in the buffers list
891 * hanging off the address_space structure.
892 */
893int generic_file_fsync(struct file *file, int datasync)
d5aacad5
AV
894{
895 struct writeback_control wbc = {
896 .sync_mode = WB_SYNC_ALL,
897 .nr_to_write = 0, /* metadata-only; caller takes care of data */
898 };
7ea80859 899 struct inode *inode = file->f_mapping->host;
d5aacad5
AV
900 int err;
901 int ret;
902
903 ret = sync_mapping_buffers(inode->i_mapping);
904 if (!(inode->i_state & I_DIRTY))
905 return ret;
906 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
907 return ret;
908
909 err = sync_inode(inode, &wbc);
910 if (ret == 0)
911 ret = err;
912 return ret;
913}
1b061d92
CH
914EXPORT_SYMBOL(generic_file_fsync);
915
30ca22c7
PL
916/**
917 * generic_check_addressable - Check addressability of file system
918 * @blocksize_bits: log of file system block size
919 * @num_blocks: number of blocks in file system
920 *
921 * Determine whether a file system with @num_blocks blocks (and a
922 * block size of 2**@blocksize_bits) is addressable by the sector_t
923 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
924 */
925int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
926{
927 u64 last_fs_block = num_blocks - 1;
928
929 if (unlikely(num_blocks == 0))
930 return 0;
931
932 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
933 return -EINVAL;
934
935 if ((last_fs_block >
936 (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
937 (last_fs_block >
938 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - blocksize_bits))) {
939 return -EFBIG;
940 }
941 return 0;
942}
943EXPORT_SYMBOL(generic_check_addressable);
944
1b061d92
CH
945/*
946 * No-op implementation of ->fsync for in-memory filesystems.
947 */
948int noop_fsync(struct file *file, int datasync)
949{
950 return 0;
951}
d5aacad5 952
1da177e4
LT
953EXPORT_SYMBOL(dcache_dir_close);
954EXPORT_SYMBOL(dcache_dir_lseek);
955EXPORT_SYMBOL(dcache_dir_open);
956EXPORT_SYMBOL(dcache_readdir);
957EXPORT_SYMBOL(generic_read_dir);
958EXPORT_SYMBOL(get_sb_pseudo);
afddba49
NP
959EXPORT_SYMBOL(simple_write_begin);
960EXPORT_SYMBOL(simple_write_end);
1da177e4
LT
961EXPORT_SYMBOL(simple_dir_inode_operations);
962EXPORT_SYMBOL(simple_dir_operations);
963EXPORT_SYMBOL(simple_empty);
1da177e4
LT
964EXPORT_SYMBOL(simple_fill_super);
965EXPORT_SYMBOL(simple_getattr);
966EXPORT_SYMBOL(simple_link);
967EXPORT_SYMBOL(simple_lookup);
968EXPORT_SYMBOL(simple_pin_fs);
1da177e4
LT
969EXPORT_SYMBOL(simple_readpage);
970EXPORT_SYMBOL(simple_release_fs);
971EXPORT_SYMBOL(simple_rename);
972EXPORT_SYMBOL(simple_rmdir);
973EXPORT_SYMBOL(simple_statfs);
1b061d92 974EXPORT_SYMBOL(noop_fsync);
1da177e4
LT
975EXPORT_SYMBOL(simple_unlink);
976EXPORT_SYMBOL(simple_read_from_buffer);
6a727b43 977EXPORT_SYMBOL(simple_write_to_buffer);
93b07113 978EXPORT_SYMBOL(memory_read_from_buffer);
76791ab2 979EXPORT_SYMBOL(simple_transaction_set);
1da177e4
LT
980EXPORT_SYMBOL(simple_transaction_get);
981EXPORT_SYMBOL(simple_transaction_read);
982EXPORT_SYMBOL(simple_transaction_release);
acaefc25 983EXPORT_SYMBOL_GPL(simple_attr_open);
74bedc4d 984EXPORT_SYMBOL_GPL(simple_attr_release);
acaefc25
AB
985EXPORT_SYMBOL_GPL(simple_attr_read);
986EXPORT_SYMBOL_GPL(simple_attr_write);