]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/libfs.c
vfs: push dentry_unhash on rename_dir into file systems
[mirror_ubuntu-artful-kernel.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
da502956
NP
19static inline int simple_positive(struct dentry *dentry)
20{
21 return dentry->d_inode && !d_unhashed(dentry);
22}
23
1da177e4
LT
24int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
25 struct kstat *stat)
26{
27 struct inode *inode = dentry->d_inode;
28 generic_fillattr(inode, stat);
29 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
30 return 0;
31}
32
726c3342 33int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
1da177e4 34{
726c3342 35 buf->f_type = dentry->d_sb->s_magic;
1da177e4
LT
36 buf->f_bsize = PAGE_CACHE_SIZE;
37 buf->f_namelen = NAME_MAX;
38 return 0;
39}
40
41/*
42 * Retaining negative dentries for an in-memory filesystem just wastes
43 * memory and lookup time: arrange for them to be deleted immediately.
44 */
fe15ce44 45static int simple_delete_dentry(const struct dentry *dentry)
1da177e4
LT
46{
47 return 1;
48}
49
50/*
51 * Lookup the data. This is trivial - if the dentry didn't already
52 * exist, we know it is negative. Set d_op to delete negative dentries.
53 */
54struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
55{
3ba13d17 56 static const struct dentry_operations simple_dentry_operations = {
1da177e4
LT
57 .d_delete = simple_delete_dentry,
58 };
59
60 if (dentry->d_name.len > NAME_MAX)
61 return ERR_PTR(-ENAMETOOLONG);
fb045adb 62 d_set_d_op(dentry, &simple_dentry_operations);
1da177e4
LT
63 d_add(dentry, NULL);
64 return NULL;
65}
66
1da177e4
LT
67int dcache_dir_open(struct inode *inode, struct file *file)
68{
69 static struct qstr cursor_name = {.len = 1, .name = "."};
70
0f7fc9e4 71 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
1da177e4
LT
72
73 return file->private_data ? 0 : -ENOMEM;
74}
75
76int dcache_dir_close(struct inode *inode, struct file *file)
77{
78 dput(file->private_data);
79 return 0;
80}
81
82loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
83{
2fd6b7f5
NP
84 struct dentry *dentry = file->f_path.dentry;
85 mutex_lock(&dentry->d_inode->i_mutex);
1da177e4
LT
86 switch (origin) {
87 case 1:
88 offset += file->f_pos;
89 case 0:
90 if (offset >= 0)
91 break;
92 default:
2fd6b7f5 93 mutex_unlock(&dentry->d_inode->i_mutex);
1da177e4
LT
94 return -EINVAL;
95 }
96 if (offset != file->f_pos) {
97 file->f_pos = offset;
98 if (file->f_pos >= 2) {
99 struct list_head *p;
100 struct dentry *cursor = file->private_data;
101 loff_t n = file->f_pos - 2;
102
2fd6b7f5
NP
103 spin_lock(&dentry->d_lock);
104 /* d_lock not required for cursor */
5160ee6f 105 list_del(&cursor->d_u.d_child);
2fd6b7f5
NP
106 p = dentry->d_subdirs.next;
107 while (n && p != &dentry->d_subdirs) {
1da177e4 108 struct dentry *next;
5160ee6f 109 next = list_entry(p, struct dentry, d_u.d_child);
2fd6b7f5 110 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
da502956 111 if (simple_positive(next))
1da177e4 112 n--;
da502956 113 spin_unlock(&next->d_lock);
1da177e4
LT
114 p = p->next;
115 }
5160ee6f 116 list_add_tail(&cursor->d_u.d_child, p);
2fd6b7f5 117 spin_unlock(&dentry->d_lock);
1da177e4
LT
118 }
119 }
2fd6b7f5 120 mutex_unlock(&dentry->d_inode->i_mutex);
1da177e4
LT
121 return offset;
122}
123
124/* Relationship between i_mode and the DT_xxx types */
125static inline unsigned char dt_type(struct inode *inode)
126{
127 return (inode->i_mode >> 12) & 15;
128}
129
130/*
131 * Directory is locked and all positive dentries in it are safe, since
132 * for ramfs-type trees they can't go away without unlink() or rmdir(),
133 * both impossible due to the lock on directory.
134 */
135
136int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
137{
0f7fc9e4 138 struct dentry *dentry = filp->f_path.dentry;
1da177e4 139 struct dentry *cursor = filp->private_data;
5160ee6f 140 struct list_head *p, *q = &cursor->d_u.d_child;
1da177e4
LT
141 ino_t ino;
142 int i = filp->f_pos;
143
144 switch (i) {
145 case 0:
146 ino = dentry->d_inode->i_ino;
147 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
148 break;
149 filp->f_pos++;
150 i++;
151 /* fallthrough */
152 case 1:
153 ino = parent_ino(dentry);
154 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
155 break;
156 filp->f_pos++;
157 i++;
158 /* fallthrough */
159 default:
2fd6b7f5 160 spin_lock(&dentry->d_lock);
1bfba4e8
AM
161 if (filp->f_pos == 2)
162 list_move(q, &dentry->d_subdirs);
163
1da177e4
LT
164 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
165 struct dentry *next;
5160ee6f 166 next = list_entry(p, struct dentry, d_u.d_child);
da502956
NP
167 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
168 if (!simple_positive(next)) {
169 spin_unlock(&next->d_lock);
1da177e4 170 continue;
da502956 171 }
1da177e4 172
da502956 173 spin_unlock(&next->d_lock);
2fd6b7f5 174 spin_unlock(&dentry->d_lock);
0f8952c2
RN
175 if (filldir(dirent, next->d_name.name,
176 next->d_name.len, filp->f_pos,
177 next->d_inode->i_ino,
178 dt_type(next->d_inode)) < 0)
1da177e4 179 return 0;
2fd6b7f5
NP
180 spin_lock(&dentry->d_lock);
181 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
1da177e4 182 /* next is still alive */
1bfba4e8 183 list_move(q, p);
2fd6b7f5 184 spin_unlock(&next->d_lock);
1da177e4
LT
185 p = q;
186 filp->f_pos++;
187 }
2fd6b7f5 188 spin_unlock(&dentry->d_lock);
1da177e4
LT
189 }
190 return 0;
191}
192
193ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
194{
195 return -EISDIR;
196}
197
4b6f5d20 198const struct file_operations simple_dir_operations = {
1da177e4
LT
199 .open = dcache_dir_open,
200 .release = dcache_dir_close,
201 .llseek = dcache_dir_lseek,
202 .read = generic_read_dir,
203 .readdir = dcache_readdir,
1b061d92 204 .fsync = noop_fsync,
1da177e4
LT
205};
206
92e1d5be 207const struct inode_operations simple_dir_inode_operations = {
1da177e4
LT
208 .lookup = simple_lookup,
209};
210
759b9775
HD
211static const struct super_operations simple_super_operations = {
212 .statfs = simple_statfs,
213};
214
1da177e4
LT
215/*
216 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
217 * will never be mountable)
218 */
51139ada 219struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
c74a1cbb
AV
220 const struct super_operations *ops,
221 const struct dentry_operations *dops, unsigned long magic)
1da177e4
LT
222{
223 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
1da177e4
LT
224 struct dentry *dentry;
225 struct inode *root;
226 struct qstr d_name = {.name = name, .len = strlen(name)};
227
228 if (IS_ERR(s))
51139ada 229 return ERR_CAST(s);
1da177e4
LT
230
231 s->s_flags = MS_NOUSER;
89a4eb4b 232 s->s_maxbytes = MAX_LFS_FILESIZE;
3971e1a9
AN
233 s->s_blocksize = PAGE_SIZE;
234 s->s_blocksize_bits = PAGE_SHIFT;
1da177e4 235 s->s_magic = magic;
759b9775 236 s->s_op = ops ? ops : &simple_super_operations;
1da177e4
LT
237 s->s_time_gran = 1;
238 root = new_inode(s);
239 if (!root)
240 goto Enomem;
1a1c9bb4
JL
241 /*
242 * since this is the first inode, make it number 1. New inodes created
243 * after this must take care not to collide with it (by passing
244 * max_reserved of 1 to iunique).
245 */
246 root->i_ino = 1;
1da177e4 247 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
1da177e4
LT
248 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
249 dentry = d_alloc(NULL, &d_name);
250 if (!dentry) {
251 iput(root);
252 goto Enomem;
253 }
254 dentry->d_sb = s;
255 dentry->d_parent = dentry;
256 d_instantiate(dentry, root);
257 s->s_root = dentry;
c74a1cbb 258 s->s_d_op = dops;
1da177e4 259 s->s_flags |= MS_ACTIVE;
51139ada 260 return dget(s->s_root);
1da177e4
LT
261
262Enomem:
6f5bbff9 263 deactivate_locked_super(s);
51139ada 264 return ERR_PTR(-ENOMEM);
1da177e4
LT
265}
266
267int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
268{
269 struct inode *inode = old_dentry->d_inode;
270
271 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
d8c76e6f 272 inc_nlink(inode);
7de9c6ee 273 ihold(inode);
1da177e4
LT
274 dget(dentry);
275 d_instantiate(dentry, inode);
276 return 0;
277}
278
1da177e4
LT
279int simple_empty(struct dentry *dentry)
280{
281 struct dentry *child;
282 int ret = 0;
283
2fd6b7f5 284 spin_lock(&dentry->d_lock);
da502956
NP
285 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
286 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
287 if (simple_positive(child)) {
288 spin_unlock(&child->d_lock);
1da177e4 289 goto out;
da502956
NP
290 }
291 spin_unlock(&child->d_lock);
292 }
1da177e4
LT
293 ret = 1;
294out:
2fd6b7f5 295 spin_unlock(&dentry->d_lock);
1da177e4
LT
296 return ret;
297}
298
299int simple_unlink(struct inode *dir, struct dentry *dentry)
300{
301 struct inode *inode = dentry->d_inode;
302
303 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
9a53c3a7 304 drop_nlink(inode);
1da177e4
LT
305 dput(dentry);
306 return 0;
307}
308
309int simple_rmdir(struct inode *dir, struct dentry *dentry)
310{
311 if (!simple_empty(dentry))
312 return -ENOTEMPTY;
313
79bf7c73
SW
314 dentry_unhash(dentry);
315
9a53c3a7 316 drop_nlink(dentry->d_inode);
1da177e4 317 simple_unlink(dir, dentry);
9a53c3a7 318 drop_nlink(dir);
1da177e4
LT
319 return 0;
320}
321
322int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
323 struct inode *new_dir, struct dentry *new_dentry)
324{
325 struct inode *inode = old_dentry->d_inode;
326 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
327
e4eaac06
SW
328 if (new_dentry->d_inode && S_ISDIR(new_dentry->d_inode->i_mode))
329 dentry_unhash(new_dentry);
330
1da177e4
LT
331 if (!simple_empty(new_dentry))
332 return -ENOTEMPTY;
333
334 if (new_dentry->d_inode) {
335 simple_unlink(new_dir, new_dentry);
336 if (they_are_dirs)
9a53c3a7 337 drop_nlink(old_dir);
1da177e4 338 } else if (they_are_dirs) {
9a53c3a7 339 drop_nlink(old_dir);
d8c76e6f 340 inc_nlink(new_dir);
1da177e4
LT
341 }
342
343 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
344 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
345
346 return 0;
347}
348
7bb46a67 349/**
eef2380c 350 * simple_setattr - setattr for simple filesystem
7bb46a67 351 * @dentry: dentry
352 * @iattr: iattr structure
353 *
354 * Returns 0 on success, -error on failure.
355 *
eef2380c
CH
356 * simple_setattr is a simple ->setattr implementation without a proper
357 * implementation of size changes.
358 *
359 * It can either be used for in-memory filesystems or special files
360 * on simple regular filesystems. Anything that needs to change on-disk
361 * or wire state on size changes needs its own setattr method.
7bb46a67 362 */
363int simple_setattr(struct dentry *dentry, struct iattr *iattr)
364{
365 struct inode *inode = dentry->d_inode;
366 int error;
367
eef2380c
CH
368 WARN_ON_ONCE(inode->i_op->truncate);
369
7bb46a67 370 error = inode_change_ok(inode, iattr);
371 if (error)
372 return error;
373
2c27c65e
CH
374 if (iattr->ia_valid & ATTR_SIZE)
375 truncate_setsize(inode, iattr->ia_size);
6a1a90ad 376 setattr_copy(inode, iattr);
eef2380c
CH
377 mark_inode_dirty(inode);
378 return 0;
7bb46a67 379}
380EXPORT_SYMBOL(simple_setattr);
381
1da177e4
LT
382int simple_readpage(struct file *file, struct page *page)
383{
c0d92cbc 384 clear_highpage(page);
1da177e4
LT
385 flush_dcache_page(page);
386 SetPageUptodate(page);
1da177e4
LT
387 unlock_page(page);
388 return 0;
389}
390
afddba49
NP
391int simple_write_begin(struct file *file, struct address_space *mapping,
392 loff_t pos, unsigned len, unsigned flags,
393 struct page **pagep, void **fsdata)
394{
395 struct page *page;
396 pgoff_t index;
afddba49
NP
397
398 index = pos >> PAGE_CACHE_SHIFT;
afddba49 399
54566b2c 400 page = grab_cache_page_write_begin(mapping, index, flags);
afddba49
NP
401 if (!page)
402 return -ENOMEM;
403
404 *pagep = page;
405
193cf4b9
BH
406 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
407 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
408
409 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
410 }
411 return 0;
afddba49
NP
412}
413
ad2a722f
BH
414/**
415 * simple_write_end - .write_end helper for non-block-device FSes
416 * @available: See .write_end of address_space_operations
417 * @file: "
418 * @mapping: "
419 * @pos: "
420 * @len: "
421 * @copied: "
422 * @page: "
423 * @fsdata: "
424 *
425 * simple_write_end does the minimum needed for updating a page after writing is
426 * done. It has the same API signature as the .write_end of
427 * address_space_operations vector. So it can just be set onto .write_end for
428 * FSes that don't need any other processing. i_mutex is assumed to be held.
429 * Block based filesystems should use generic_write_end().
430 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
431 * is not called, so a filesystem that actually does store data in .write_inode
432 * should extend on what's done here with a call to mark_inode_dirty() in the
433 * case that i_size has changed.
434 */
afddba49
NP
435int simple_write_end(struct file *file, struct address_space *mapping,
436 loff_t pos, unsigned len, unsigned copied,
437 struct page *page, void *fsdata)
438{
ad2a722f
BH
439 struct inode *inode = page->mapping->host;
440 loff_t last_pos = pos + copied;
afddba49
NP
441
442 /* zero the stale part of the page if we did a short copy */
443 if (copied < len) {
ad2a722f
BH
444 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
445
446 zero_user(page, from + copied, len - copied);
afddba49
NP
447 }
448
ad2a722f
BH
449 if (!PageUptodate(page))
450 SetPageUptodate(page);
451 /*
452 * No need to use i_size_read() here, the i_size
453 * cannot change under us because we hold the i_mutex.
454 */
455 if (last_pos > inode->i_size)
456 i_size_write(inode, last_pos);
afddba49 457
ad2a722f 458 set_page_dirty(page);
afddba49
NP
459 unlock_page(page);
460 page_cache_release(page);
461
462 return copied;
463}
464
1a1c9bb4
JL
465/*
466 * the inodes created here are not hashed. If you use iunique to generate
467 * unique inode values later for this filesystem, then you must take care
468 * to pass it an appropriate max_reserved value to avoid collisions.
469 */
7d683a09
RS
470int simple_fill_super(struct super_block *s, unsigned long magic,
471 struct tree_descr *files)
1da177e4 472{
1da177e4
LT
473 struct inode *inode;
474 struct dentry *root;
475 struct dentry *dentry;
476 int i;
477
478 s->s_blocksize = PAGE_CACHE_SIZE;
479 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
480 s->s_magic = magic;
759b9775 481 s->s_op = &simple_super_operations;
1da177e4
LT
482 s->s_time_gran = 1;
483
484 inode = new_inode(s);
485 if (!inode)
486 return -ENOMEM;
1a1c9bb4
JL
487 /*
488 * because the root inode is 1, the files array must not contain an
489 * entry at index 1
490 */
491 inode->i_ino = 1;
1da177e4 492 inode->i_mode = S_IFDIR | 0755;
1da177e4
LT
493 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
494 inode->i_op = &simple_dir_inode_operations;
495 inode->i_fop = &simple_dir_operations;
7656f328 496 inode->i_nlink = 2;
1da177e4
LT
497 root = d_alloc_root(inode);
498 if (!root) {
499 iput(inode);
500 return -ENOMEM;
501 }
502 for (i = 0; !files->name || files->name[0]; i++, files++) {
503 if (!files->name)
504 continue;
1a1c9bb4
JL
505
506 /* warn if it tries to conflict with the root inode */
507 if (unlikely(i == 1))
508 printk(KERN_WARNING "%s: %s passed in a files array"
509 "with an index of 1!\n", __func__,
510 s->s_type->name);
511
1da177e4
LT
512 dentry = d_alloc_name(root, files->name);
513 if (!dentry)
514 goto out;
515 inode = new_inode(s);
516 if (!inode)
517 goto out;
518 inode->i_mode = S_IFREG | files->mode;
1da177e4
LT
519 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
520 inode->i_fop = files->ops;
521 inode->i_ino = i;
522 d_add(dentry, inode);
523 }
524 s->s_root = root;
525 return 0;
526out:
527 d_genocide(root);
528 dput(root);
529 return -ENOMEM;
530}
531
532static DEFINE_SPINLOCK(pin_fs_lock);
533
1f5ce9e9 534int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
1da177e4
LT
535{
536 struct vfsmount *mnt = NULL;
537 spin_lock(&pin_fs_lock);
538 if (unlikely(!*mount)) {
539 spin_unlock(&pin_fs_lock);
1f5ce9e9 540 mnt = vfs_kern_mount(type, 0, type->name, NULL);
1da177e4
LT
541 if (IS_ERR(mnt))
542 return PTR_ERR(mnt);
543 spin_lock(&pin_fs_lock);
544 if (!*mount)
545 *mount = mnt;
546 }
547 mntget(*mount);
548 ++*count;
549 spin_unlock(&pin_fs_lock);
550 mntput(mnt);
551 return 0;
552}
553
554void simple_release_fs(struct vfsmount **mount, int *count)
555{
556 struct vfsmount *mnt;
557 spin_lock(&pin_fs_lock);
558 mnt = *mount;
559 if (!--*count)
560 *mount = NULL;
561 spin_unlock(&pin_fs_lock);
562 mntput(mnt);
563}
564
6d1029b5
AM
565/**
566 * simple_read_from_buffer - copy data from the buffer to user space
567 * @to: the user space buffer to read to
568 * @count: the maximum number of bytes to read
569 * @ppos: the current position in the buffer
570 * @from: the buffer to read from
571 * @available: the size of the buffer
572 *
573 * The simple_read_from_buffer() function reads up to @count bytes from the
574 * buffer @from at offset @ppos into the user space address starting at @to.
575 *
576 * On success, the number of bytes read is returned and the offset @ppos is
577 * advanced by this number, or negative value is returned on error.
578 **/
1da177e4
LT
579ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
580 const void *from, size_t available)
581{
582 loff_t pos = *ppos;
14be2746
SR
583 size_t ret;
584
1da177e4
LT
585 if (pos < 0)
586 return -EINVAL;
14be2746 587 if (pos >= available || !count)
1da177e4
LT
588 return 0;
589 if (count > available - pos)
590 count = available - pos;
14be2746
SR
591 ret = copy_to_user(to, from + pos, count);
592 if (ret == count)
1da177e4 593 return -EFAULT;
14be2746 594 count -= ret;
1da177e4
LT
595 *ppos = pos + count;
596 return count;
597}
598
6a727b43
JS
599/**
600 * simple_write_to_buffer - copy data from user space to the buffer
601 * @to: the buffer to write to
602 * @available: the size of the buffer
603 * @ppos: the current position in the buffer
604 * @from: the user space buffer to read from
605 * @count: the maximum number of bytes to read
606 *
607 * The simple_write_to_buffer() function reads up to @count bytes from the user
608 * space address starting at @from into the buffer @to at offset @ppos.
609 *
610 * On success, the number of bytes written is returned and the offset @ppos is
611 * advanced by this number, or negative value is returned on error.
612 **/
613ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
614 const void __user *from, size_t count)
615{
616 loff_t pos = *ppos;
617 size_t res;
618
619 if (pos < 0)
620 return -EINVAL;
621 if (pos >= available || !count)
622 return 0;
623 if (count > available - pos)
624 count = available - pos;
625 res = copy_from_user(to + pos, from, count);
626 if (res == count)
627 return -EFAULT;
628 count -= res;
629 *ppos = pos + count;
630 return count;
631}
632
6d1029b5
AM
633/**
634 * memory_read_from_buffer - copy data from the buffer
635 * @to: the kernel space buffer to read to
636 * @count: the maximum number of bytes to read
637 * @ppos: the current position in the buffer
638 * @from: the buffer to read from
639 * @available: the size of the buffer
640 *
641 * The memory_read_from_buffer() function reads up to @count bytes from the
642 * buffer @from at offset @ppos into the kernel space address starting at @to.
643 *
644 * On success, the number of bytes read is returned and the offset @ppos is
645 * advanced by this number, or negative value is returned on error.
646 **/
93b07113
AM
647ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
648 const void *from, size_t available)
649{
650 loff_t pos = *ppos;
651
652 if (pos < 0)
653 return -EINVAL;
654 if (pos >= available)
655 return 0;
656 if (count > available - pos)
657 count = available - pos;
658 memcpy(to, from + pos, count);
659 *ppos = pos + count;
660
661 return count;
662}
663
1da177e4
LT
664/*
665 * Transaction based IO.
666 * The file expects a single write which triggers the transaction, and then
667 * possibly a read which collects the result - which is stored in a
668 * file-local buffer.
669 */
76791ab2
IM
670
671void simple_transaction_set(struct file *file, size_t n)
672{
673 struct simple_transaction_argresp *ar = file->private_data;
674
675 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
676
677 /*
678 * The barrier ensures that ar->size will really remain zero until
679 * ar->data is ready for reading.
680 */
681 smp_mb();
682 ar->size = n;
683}
684
1da177e4
LT
685char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
686{
687 struct simple_transaction_argresp *ar;
688 static DEFINE_SPINLOCK(simple_transaction_lock);
689
690 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
691 return ERR_PTR(-EFBIG);
692
693 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
694 if (!ar)
695 return ERR_PTR(-ENOMEM);
696
697 spin_lock(&simple_transaction_lock);
698
699 /* only one write allowed per open */
700 if (file->private_data) {
701 spin_unlock(&simple_transaction_lock);
702 free_page((unsigned long)ar);
703 return ERR_PTR(-EBUSY);
704 }
705
706 file->private_data = ar;
707
708 spin_unlock(&simple_transaction_lock);
709
710 if (copy_from_user(ar->data, buf, size))
711 return ERR_PTR(-EFAULT);
712
713 return ar->data;
714}
715
716ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
717{
718 struct simple_transaction_argresp *ar = file->private_data;
719
720 if (!ar)
721 return 0;
722 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
723}
724
725int simple_transaction_release(struct inode *inode, struct file *file)
726{
727 free_page((unsigned long)file->private_data);
728 return 0;
729}
730
acaefc25
AB
731/* Simple attribute files */
732
733struct simple_attr {
8b88b099
CH
734 int (*get)(void *, u64 *);
735 int (*set)(void *, u64);
acaefc25
AB
736 char get_buf[24]; /* enough to store a u64 and "\n\0" */
737 char set_buf[24];
738 void *data;
739 const char *fmt; /* format for read operation */
7cf34c76 740 struct mutex mutex; /* protects access to these buffers */
acaefc25
AB
741};
742
743/* simple_attr_open is called by an actual attribute open file operation
744 * to set the attribute specific access operations. */
745int simple_attr_open(struct inode *inode, struct file *file,
8b88b099 746 int (*get)(void *, u64 *), int (*set)(void *, u64),
acaefc25
AB
747 const char *fmt)
748{
749 struct simple_attr *attr;
750
751 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
752 if (!attr)
753 return -ENOMEM;
754
755 attr->get = get;
756 attr->set = set;
8e18e294 757 attr->data = inode->i_private;
acaefc25 758 attr->fmt = fmt;
7cf34c76 759 mutex_init(&attr->mutex);
acaefc25
AB
760
761 file->private_data = attr;
762
763 return nonseekable_open(inode, file);
764}
765
74bedc4d 766int simple_attr_release(struct inode *inode, struct file *file)
acaefc25
AB
767{
768 kfree(file->private_data);
769 return 0;
770}
771
772/* read from the buffer that is filled with the get function */
773ssize_t simple_attr_read(struct file *file, char __user *buf,
774 size_t len, loff_t *ppos)
775{
776 struct simple_attr *attr;
777 size_t size;
778 ssize_t ret;
779
780 attr = file->private_data;
781
782 if (!attr->get)
783 return -EACCES;
784
9261303a
CH
785 ret = mutex_lock_interruptible(&attr->mutex);
786 if (ret)
787 return ret;
788
8b88b099 789 if (*ppos) { /* continued read */
acaefc25 790 size = strlen(attr->get_buf);
8b88b099
CH
791 } else { /* first read */
792 u64 val;
793 ret = attr->get(attr->data, &val);
794 if (ret)
795 goto out;
796
acaefc25 797 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
8b88b099
CH
798 attr->fmt, (unsigned long long)val);
799 }
acaefc25
AB
800
801 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
8b88b099 802out:
7cf34c76 803 mutex_unlock(&attr->mutex);
acaefc25
AB
804 return ret;
805}
806
807/* interpret the buffer as a number to call the set function with */
808ssize_t simple_attr_write(struct file *file, const char __user *buf,
809 size_t len, loff_t *ppos)
810{
811 struct simple_attr *attr;
812 u64 val;
813 size_t size;
814 ssize_t ret;
815
816 attr = file->private_data;
acaefc25
AB
817 if (!attr->set)
818 return -EACCES;
819
9261303a
CH
820 ret = mutex_lock_interruptible(&attr->mutex);
821 if (ret)
822 return ret;
823
acaefc25
AB
824 ret = -EFAULT;
825 size = min(sizeof(attr->set_buf) - 1, len);
826 if (copy_from_user(attr->set_buf, buf, size))
827 goto out;
828
acaefc25
AB
829 attr->set_buf[size] = '\0';
830 val = simple_strtol(attr->set_buf, NULL, 0);
05cc0cee
WF
831 ret = attr->set(attr->data, val);
832 if (ret == 0)
833 ret = len; /* on success, claim we got the whole input */
acaefc25 834out:
7cf34c76 835 mutex_unlock(&attr->mutex);
acaefc25
AB
836 return ret;
837}
838
2596110a
CH
839/**
840 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
841 * @sb: filesystem to do the file handle conversion on
842 * @fid: file handle to convert
843 * @fh_len: length of the file handle in bytes
844 * @fh_type: type of file handle
845 * @get_inode: filesystem callback to retrieve inode
846 *
847 * This function decodes @fid as long as it has one of the well-known
848 * Linux filehandle types and calls @get_inode on it to retrieve the
849 * inode for the object specified in the file handle.
850 */
851struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
852 int fh_len, int fh_type, struct inode *(*get_inode)
853 (struct super_block *sb, u64 ino, u32 gen))
854{
855 struct inode *inode = NULL;
856
857 if (fh_len < 2)
858 return NULL;
859
860 switch (fh_type) {
861 case FILEID_INO32_GEN:
862 case FILEID_INO32_GEN_PARENT:
863 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
864 break;
865 }
866
4ea3ada2 867 return d_obtain_alias(inode);
2596110a
CH
868}
869EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
870
871/**
872 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
873 * @sb: filesystem to do the file handle conversion on
874 * @fid: file handle to convert
875 * @fh_len: length of the file handle in bytes
876 * @fh_type: type of file handle
877 * @get_inode: filesystem callback to retrieve inode
878 *
879 * This function decodes @fid as long as it has one of the well-known
880 * Linux filehandle types and calls @get_inode on it to retrieve the
881 * inode for the _parent_ object specified in the file handle if it
882 * is specified in the file handle, or NULL otherwise.
883 */
884struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
885 int fh_len, int fh_type, struct inode *(*get_inode)
886 (struct super_block *sb, u64 ino, u32 gen))
887{
888 struct inode *inode = NULL;
889
890 if (fh_len <= 2)
891 return NULL;
892
893 switch (fh_type) {
894 case FILEID_INO32_GEN_PARENT:
895 inode = get_inode(sb, fid->i32.parent_ino,
896 (fh_len > 3 ? fid->i32.parent_gen : 0));
897 break;
898 }
899
4ea3ada2 900 return d_obtain_alias(inode);
2596110a
CH
901}
902EXPORT_SYMBOL_GPL(generic_fh_to_parent);
903
1b061d92
CH
904/**
905 * generic_file_fsync - generic fsync implementation for simple filesystems
906 * @file: file to synchronize
907 * @datasync: only synchronize essential metadata if true
908 *
909 * This is a generic implementation of the fsync method for simple
910 * filesystems which track all non-inode metadata in the buffers list
911 * hanging off the address_space structure.
912 */
913int generic_file_fsync(struct file *file, int datasync)
d5aacad5 914{
7ea80859 915 struct inode *inode = file->f_mapping->host;
d5aacad5
AV
916 int err;
917 int ret;
918
919 ret = sync_mapping_buffers(inode->i_mapping);
920 if (!(inode->i_state & I_DIRTY))
921 return ret;
922 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
923 return ret;
924
c3765016 925 err = sync_inode_metadata(inode, 1);
d5aacad5
AV
926 if (ret == 0)
927 ret = err;
928 return ret;
929}
1b061d92
CH
930EXPORT_SYMBOL(generic_file_fsync);
931
30ca22c7
PL
932/**
933 * generic_check_addressable - Check addressability of file system
934 * @blocksize_bits: log of file system block size
935 * @num_blocks: number of blocks in file system
936 *
937 * Determine whether a file system with @num_blocks blocks (and a
938 * block size of 2**@blocksize_bits) is addressable by the sector_t
939 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
940 */
941int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
942{
943 u64 last_fs_block = num_blocks - 1;
a33f13ef
JB
944 u64 last_fs_page =
945 last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
30ca22c7
PL
946
947 if (unlikely(num_blocks == 0))
948 return 0;
949
950 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
951 return -EINVAL;
952
a33f13ef
JB
953 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
954 (last_fs_page > (pgoff_t)(~0ULL))) {
30ca22c7
PL
955 return -EFBIG;
956 }
957 return 0;
958}
959EXPORT_SYMBOL(generic_check_addressable);
960
1b061d92
CH
961/*
962 * No-op implementation of ->fsync for in-memory filesystems.
963 */
964int noop_fsync(struct file *file, int datasync)
965{
966 return 0;
967}
d5aacad5 968
1da177e4
LT
969EXPORT_SYMBOL(dcache_dir_close);
970EXPORT_SYMBOL(dcache_dir_lseek);
971EXPORT_SYMBOL(dcache_dir_open);
972EXPORT_SYMBOL(dcache_readdir);
973EXPORT_SYMBOL(generic_read_dir);
51139ada 974EXPORT_SYMBOL(mount_pseudo);
afddba49
NP
975EXPORT_SYMBOL(simple_write_begin);
976EXPORT_SYMBOL(simple_write_end);
1da177e4
LT
977EXPORT_SYMBOL(simple_dir_inode_operations);
978EXPORT_SYMBOL(simple_dir_operations);
979EXPORT_SYMBOL(simple_empty);
1da177e4
LT
980EXPORT_SYMBOL(simple_fill_super);
981EXPORT_SYMBOL(simple_getattr);
982EXPORT_SYMBOL(simple_link);
983EXPORT_SYMBOL(simple_lookup);
984EXPORT_SYMBOL(simple_pin_fs);
1da177e4
LT
985EXPORT_SYMBOL(simple_readpage);
986EXPORT_SYMBOL(simple_release_fs);
987EXPORT_SYMBOL(simple_rename);
988EXPORT_SYMBOL(simple_rmdir);
989EXPORT_SYMBOL(simple_statfs);
1b061d92 990EXPORT_SYMBOL(noop_fsync);
1da177e4
LT
991EXPORT_SYMBOL(simple_unlink);
992EXPORT_SYMBOL(simple_read_from_buffer);
6a727b43 993EXPORT_SYMBOL(simple_write_to_buffer);
93b07113 994EXPORT_SYMBOL(memory_read_from_buffer);
76791ab2 995EXPORT_SYMBOL(simple_transaction_set);
1da177e4
LT
996EXPORT_SYMBOL(simple_transaction_get);
997EXPORT_SYMBOL(simple_transaction_read);
998EXPORT_SYMBOL(simple_transaction_release);
acaefc25 999EXPORT_SYMBOL_GPL(simple_attr_open);
74bedc4d 1000EXPORT_SYMBOL_GPL(simple_attr_release);
acaefc25
AB
1001EXPORT_SYMBOL_GPL(simple_attr_read);
1002EXPORT_SYMBOL_GPL(simple_attr_write);