blob: 232bea425b09155a1f116981e4e53283a1738bbc [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * fs/libfs.c
3 * Library for filesystems writers.
4 */
5
6#include <linux/module.h>
7#include <linux/pagemap.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +09008#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/mount.h>
10#include <linux/vfs.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080011#include <linux/mutex.h>
Christoph Hellwig25961102007-10-21 16:42:05 -070012#include <linux/exportfs.h>
Al Virod5aacad2009-06-07 14:56:44 -040013#include <linux/writeback.h>
14#include <linux/buffer_head.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080015
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <asm/uaccess.h>
17
18int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
19 struct kstat *stat)
20{
21 struct inode *inode = dentry->d_inode;
22 generic_fillattr(inode, stat);
23 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
24 return 0;
25}
26
David Howells726c3342006-06-23 02:02:58 -070027int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070028{
David Howells726c3342006-06-23 02:02:58 -070029 buf->f_type = dentry->d_sb->s_magic;
Linus Torvalds1da177e2005-04-16 15:20:36 -070030 buf->f_bsize = PAGE_CACHE_SIZE;
31 buf->f_namelen = NAME_MAX;
32 return 0;
33}
34
35/*
36 * Retaining negative dentries for an in-memory filesystem just wastes
37 * memory and lookup time: arrange for them to be deleted immediately.
38 */
39static int simple_delete_dentry(struct dentry *dentry)
40{
41 return 1;
42}
43
44/*
45 * Lookup the data. This is trivial - if the dentry didn't already
46 * exist, we know it is negative. Set d_op to delete negative dentries.
47 */
48struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
49{
Al Viro3ba13d12009-02-20 06:02:22 +000050 static const struct dentry_operations simple_dentry_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 .d_delete = simple_delete_dentry,
52 };
53
54 if (dentry->d_name.len > NAME_MAX)
55 return ERR_PTR(-ENAMETOOLONG);
56 dentry->d_op = &simple_dentry_operations;
57 d_add(dentry, NULL);
58 return NULL;
59}
60
61int simple_sync_file(struct file * file, struct dentry *dentry, int datasync)
62{
63 return 0;
64}
65
66int dcache_dir_open(struct inode *inode, struct file *file)
67{
68 static struct qstr cursor_name = {.len = 1, .name = "."};
69
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080070 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -070071
72 return file->private_data ? 0 : -ENOMEM;
73}
74
75int dcache_dir_close(struct inode *inode, struct file *file)
76{
77 dput(file->private_data);
78 return 0;
79}
80
81loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
82{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080083 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 switch (origin) {
85 case 1:
86 offset += file->f_pos;
87 case 0:
88 if (offset >= 0)
89 break;
90 default:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080091 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 return -EINVAL;
93 }
94 if (offset != file->f_pos) {
95 file->f_pos = offset;
96 if (file->f_pos >= 2) {
97 struct list_head *p;
98 struct dentry *cursor = file->private_data;
99 loff_t n = file->f_pos - 2;
100
101 spin_lock(&dcache_lock);
Eric Dumazet5160ee62006-01-08 01:03:32 -0800102 list_del(&cursor->d_u.d_child);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800103 p = file->f_path.dentry->d_subdirs.next;
104 while (n && p != &file->f_path.dentry->d_subdirs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 struct dentry *next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800106 next = list_entry(p, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 if (!d_unhashed(next) && next->d_inode)
108 n--;
109 p = p->next;
110 }
Eric Dumazet5160ee62006-01-08 01:03:32 -0800111 list_add_tail(&cursor->d_u.d_child, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 spin_unlock(&dcache_lock);
113 }
114 }
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800115 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 return offset;
117}
118
119/* Relationship between i_mode and the DT_xxx types */
120static inline unsigned char dt_type(struct inode *inode)
121{
122 return (inode->i_mode >> 12) & 15;
123}
124
125/*
126 * Directory is locked and all positive dentries in it are safe, since
127 * for ramfs-type trees they can't go away without unlink() or rmdir(),
128 * both impossible due to the lock on directory.
129 */
130
131int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
132{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800133 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 struct dentry *cursor = filp->private_data;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800135 struct list_head *p, *q = &cursor->d_u.d_child;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 ino_t ino;
137 int i = filp->f_pos;
138
139 switch (i) {
140 case 0:
141 ino = dentry->d_inode->i_ino;
142 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
143 break;
144 filp->f_pos++;
145 i++;
146 /* fallthrough */
147 case 1:
148 ino = parent_ino(dentry);
149 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
150 break;
151 filp->f_pos++;
152 i++;
153 /* fallthrough */
154 default:
155 spin_lock(&dcache_lock);
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700156 if (filp->f_pos == 2)
157 list_move(q, &dentry->d_subdirs);
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
160 struct dentry *next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800161 next = list_entry(p, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 if (d_unhashed(next) || !next->d_inode)
163 continue;
164
165 spin_unlock(&dcache_lock);
Ronni Nielsen0f8952c22007-05-09 06:44:57 +0200166 if (filldir(dirent, next->d_name.name,
167 next->d_name.len, filp->f_pos,
168 next->d_inode->i_ino,
169 dt_type(next->d_inode)) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 return 0;
171 spin_lock(&dcache_lock);
172 /* next is still alive */
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700173 list_move(q, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 p = q;
175 filp->f_pos++;
176 }
177 spin_unlock(&dcache_lock);
178 }
179 return 0;
180}
181
182ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
183{
184 return -EISDIR;
185}
186
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800187const struct file_operations simple_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 .open = dcache_dir_open,
189 .release = dcache_dir_close,
190 .llseek = dcache_dir_lseek,
191 .read = generic_read_dir,
192 .readdir = dcache_readdir,
Hugh Dickins8ae0b772005-06-25 14:55:41 -0700193 .fsync = simple_sync_file,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194};
195
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -0800196const struct inode_operations simple_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 .lookup = simple_lookup,
198};
199
Hugh Dickins759b9772007-03-05 00:30:28 -0800200static const struct super_operations simple_super_operations = {
201 .statfs = simple_statfs,
202};
203
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204/*
205 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
206 * will never be mountable)
207 */
David Howells454e2392006-06-23 02:02:57 -0700208int get_sb_pseudo(struct file_system_type *fs_type, char *name,
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800209 const struct super_operations *ops, unsigned long magic,
David Howells454e2392006-06-23 02:02:57 -0700210 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211{
212 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 struct dentry *dentry;
214 struct inode *root;
215 struct qstr d_name = {.name = name, .len = strlen(name)};
216
217 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700218 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
220 s->s_flags = MS_NOUSER;
Jeff Layton89a4eb42009-08-18 14:11:08 -0700221 s->s_maxbytes = MAX_LFS_FILESIZE;
Alex Nixon3971e1a2008-07-29 22:33:03 -0700222 s->s_blocksize = PAGE_SIZE;
223 s->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800225 s->s_op = ops ? ops : &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 s->s_time_gran = 1;
227 root = new_inode(s);
228 if (!root)
229 goto Enomem;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700230 /*
231 * since this is the first inode, make it number 1. New inodes created
232 * after this must take care not to collide with it (by passing
233 * max_reserved of 1 to iunique).
234 */
235 root->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
238 dentry = d_alloc(NULL, &d_name);
239 if (!dentry) {
240 iput(root);
241 goto Enomem;
242 }
243 dentry->d_sb = s;
244 dentry->d_parent = dentry;
245 d_instantiate(dentry, root);
246 s->s_root = dentry;
247 s->s_flags |= MS_ACTIVE;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800248 simple_set_mnt(mnt, s);
249 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250
251Enomem:
Al Viro6f5bbff2009-05-06 01:34:22 -0400252 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700253 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254}
255
256int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
257{
258 struct inode *inode = old_dentry->d_inode;
259
260 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
Dave Hansend8c76e62006-09-30 23:29:04 -0700261 inc_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 atomic_inc(&inode->i_count);
263 dget(dentry);
264 d_instantiate(dentry, inode);
265 return 0;
266}
267
268static inline int simple_positive(struct dentry *dentry)
269{
270 return dentry->d_inode && !d_unhashed(dentry);
271}
272
273int simple_empty(struct dentry *dentry)
274{
275 struct dentry *child;
276 int ret = 0;
277
278 spin_lock(&dcache_lock);
Eric Dumazet5160ee62006-01-08 01:03:32 -0800279 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 if (simple_positive(child))
281 goto out;
282 ret = 1;
283out:
284 spin_unlock(&dcache_lock);
285 return ret;
286}
287
288int simple_unlink(struct inode *dir, struct dentry *dentry)
289{
290 struct inode *inode = dentry->d_inode;
291
292 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700293 drop_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 dput(dentry);
295 return 0;
296}
297
298int simple_rmdir(struct inode *dir, struct dentry *dentry)
299{
300 if (!simple_empty(dentry))
301 return -ENOTEMPTY;
302
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700303 drop_nlink(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 simple_unlink(dir, dentry);
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700305 drop_nlink(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 return 0;
307}
308
309int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
310 struct inode *new_dir, struct dentry *new_dentry)
311{
312 struct inode *inode = old_dentry->d_inode;
313 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
314
315 if (!simple_empty(new_dentry))
316 return -ENOTEMPTY;
317
318 if (new_dentry->d_inode) {
319 simple_unlink(new_dir, new_dentry);
320 if (they_are_dirs)
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700321 drop_nlink(old_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 } else if (they_are_dirs) {
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700323 drop_nlink(old_dir);
Dave Hansend8c76e62006-09-30 23:29:04 -0700324 inc_nlink(new_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 }
326
327 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
328 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
329
330 return 0;
331}
332
333int simple_readpage(struct file *file, struct page *page)
334{
Pekka J Enbergc0d92cb2006-09-29 01:59:09 -0700335 clear_highpage(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 flush_dcache_page(page);
337 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 unlock_page(page);
339 return 0;
340}
341
Nick Pigginafddba42007-10-16 01:25:01 -0700342int simple_write_begin(struct file *file, struct address_space *mapping,
343 loff_t pos, unsigned len, unsigned flags,
344 struct page **pagep, void **fsdata)
345{
346 struct page *page;
347 pgoff_t index;
Nick Pigginafddba42007-10-16 01:25:01 -0700348
349 index = pos >> PAGE_CACHE_SHIFT;
Nick Pigginafddba42007-10-16 01:25:01 -0700350
Nick Piggin54566b22009-01-04 12:00:53 -0800351 page = grab_cache_page_write_begin(mapping, index, flags);
Nick Pigginafddba42007-10-16 01:25:01 -0700352 if (!page)
353 return -ENOMEM;
354
355 *pagep = page;
356
Boaz Harrosh193cf4b2010-01-12 16:18:08 +0200357 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
358 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
359
360 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
361 }
362 return 0;
Nick Pigginafddba42007-10-16 01:25:01 -0700363}
364
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200365/**
366 * simple_write_end - .write_end helper for non-block-device FSes
367 * @available: See .write_end of address_space_operations
368 * @file: "
369 * @mapping: "
370 * @pos: "
371 * @len: "
372 * @copied: "
373 * @page: "
374 * @fsdata: "
375 *
376 * simple_write_end does the minimum needed for updating a page after writing is
377 * done. It has the same API signature as the .write_end of
378 * address_space_operations vector. So it can just be set onto .write_end for
379 * FSes that don't need any other processing. i_mutex is assumed to be held.
380 * Block based filesystems should use generic_write_end().
381 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
382 * is not called, so a filesystem that actually does store data in .write_inode
383 * should extend on what's done here with a call to mark_inode_dirty() in the
384 * case that i_size has changed.
385 */
386int simple_write_end(struct file *file, struct address_space *mapping,
387 loff_t pos, unsigned len, unsigned copied,
388 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389{
390 struct inode *inode = page->mapping->host;
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200391 loff_t last_pos = pos + copied;
392
393 /* zero the stale part of the page if we did a short copy */
394 if (copied < len) {
395 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
396
397 zero_user(page, from + copied, len - copied);
398 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
Nick Piggin955eff52007-02-20 13:58:08 -0800400 if (!PageUptodate(page))
401 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 /*
403 * No need to use i_size_read() here, the i_size
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800404 * cannot change under us because we hold the i_mutex.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 */
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200406 if (last_pos > inode->i_size)
407 i_size_write(inode, last_pos);
408
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 set_page_dirty(page);
Nick Pigginafddba42007-10-16 01:25:01 -0700410 unlock_page(page);
411 page_cache_release(page);
412
413 return copied;
414}
415
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700416/*
417 * the inodes created here are not hashed. If you use iunique to generate
418 * unique inode values later for this filesystem, then you must take care
419 * to pass it an appropriate max_reserved value to avoid collisions.
420 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421int simple_fill_super(struct super_block *s, int magic, struct tree_descr *files)
422{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 struct inode *inode;
424 struct dentry *root;
425 struct dentry *dentry;
426 int i;
427
428 s->s_blocksize = PAGE_CACHE_SIZE;
429 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
430 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800431 s->s_op = &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 s->s_time_gran = 1;
433
434 inode = new_inode(s);
435 if (!inode)
436 return -ENOMEM;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700437 /*
438 * because the root inode is 1, the files array must not contain an
439 * entry at index 1
440 */
441 inode->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 inode->i_mode = S_IFDIR | 0755;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
444 inode->i_op = &simple_dir_inode_operations;
445 inode->i_fop = &simple_dir_operations;
Vincent Hanquez7656f322006-02-03 03:04:48 -0800446 inode->i_nlink = 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 root = d_alloc_root(inode);
448 if (!root) {
449 iput(inode);
450 return -ENOMEM;
451 }
452 for (i = 0; !files->name || files->name[0]; i++, files++) {
453 if (!files->name)
454 continue;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700455
456 /* warn if it tries to conflict with the root inode */
457 if (unlikely(i == 1))
458 printk(KERN_WARNING "%s: %s passed in a files array"
459 "with an index of 1!\n", __func__,
460 s->s_type->name);
461
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 dentry = d_alloc_name(root, files->name);
463 if (!dentry)
464 goto out;
465 inode = new_inode(s);
466 if (!inode)
467 goto out;
468 inode->i_mode = S_IFREG | files->mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
470 inode->i_fop = files->ops;
471 inode->i_ino = i;
472 d_add(dentry, inode);
473 }
474 s->s_root = root;
475 return 0;
476out:
477 d_genocide(root);
478 dput(root);
479 return -ENOMEM;
480}
481
482static DEFINE_SPINLOCK(pin_fs_lock);
483
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400484int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485{
486 struct vfsmount *mnt = NULL;
487 spin_lock(&pin_fs_lock);
488 if (unlikely(!*mount)) {
489 spin_unlock(&pin_fs_lock);
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400490 mnt = vfs_kern_mount(type, 0, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 if (IS_ERR(mnt))
492 return PTR_ERR(mnt);
493 spin_lock(&pin_fs_lock);
494 if (!*mount)
495 *mount = mnt;
496 }
497 mntget(*mount);
498 ++*count;
499 spin_unlock(&pin_fs_lock);
500 mntput(mnt);
501 return 0;
502}
503
504void simple_release_fs(struct vfsmount **mount, int *count)
505{
506 struct vfsmount *mnt;
507 spin_lock(&pin_fs_lock);
508 mnt = *mount;
509 if (!--*count)
510 *mount = NULL;
511 spin_unlock(&pin_fs_lock);
512 mntput(mnt);
513}
514
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700515/**
516 * simple_read_from_buffer - copy data from the buffer to user space
517 * @to: the user space buffer to read to
518 * @count: the maximum number of bytes to read
519 * @ppos: the current position in the buffer
520 * @from: the buffer to read from
521 * @available: the size of the buffer
522 *
523 * The simple_read_from_buffer() function reads up to @count bytes from the
524 * buffer @from at offset @ppos into the user space address starting at @to.
525 *
526 * On success, the number of bytes read is returned and the offset @ppos is
527 * advanced by this number, or negative value is returned on error.
528 **/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
530 const void *from, size_t available)
531{
532 loff_t pos = *ppos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700533 size_t ret;
534
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 if (pos < 0)
536 return -EINVAL;
Steven Rostedt14be2742009-09-18 13:05:42 -0700537 if (pos >= available || !count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 return 0;
539 if (count > available - pos)
540 count = available - pos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700541 ret = copy_to_user(to, from + pos, count);
542 if (ret == count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 return -EFAULT;
Steven Rostedt14be2742009-09-18 13:05:42 -0700544 count -= ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 *ppos = pos + count;
546 return count;
547}
548
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700549/**
Jiri Slaby6a727b42010-05-01 23:51:22 +0200550 * simple_write_to_buffer - copy data from user space to the buffer
551 * @to: the buffer to write to
552 * @available: the size of the buffer
553 * @ppos: the current position in the buffer
554 * @from: the user space buffer to read from
555 * @count: the maximum number of bytes to read
556 *
557 * The simple_write_to_buffer() function reads up to @count bytes from the user
558 * space address starting at @from into the buffer @to at offset @ppos.
559 *
560 * On success, the number of bytes written is returned and the offset @ppos is
561 * advanced by this number, or negative value is returned on error.
562 **/
563ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
564 const void __user *from, size_t count)
565{
566 loff_t pos = *ppos;
567 size_t res;
568
569 if (pos < 0)
570 return -EINVAL;
571 if (pos >= available || !count)
572 return 0;
573 if (count > available - pos)
574 count = available - pos;
575 res = copy_from_user(to + pos, from, count);
576 if (res == count)
577 return -EFAULT;
578 count -= res;
579 *ppos = pos + count;
580 return count;
581}
582
583/**
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700584 * memory_read_from_buffer - copy data from the buffer
585 * @to: the kernel space buffer to read to
586 * @count: the maximum number of bytes to read
587 * @ppos: the current position in the buffer
588 * @from: the buffer to read from
589 * @available: the size of the buffer
590 *
591 * The memory_read_from_buffer() function reads up to @count bytes from the
592 * buffer @from at offset @ppos into the kernel space address starting at @to.
593 *
594 * On success, the number of bytes read is returned and the offset @ppos is
595 * advanced by this number, or negative value is returned on error.
596 **/
Akinobu Mita93b07112008-06-05 22:46:21 -0700597ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
598 const void *from, size_t available)
599{
600 loff_t pos = *ppos;
601
602 if (pos < 0)
603 return -EINVAL;
604 if (pos >= available)
605 return 0;
606 if (count > available - pos)
607 count = available - pos;
608 memcpy(to, from + pos, count);
609 *ppos = pos + count;
610
611 return count;
612}
613
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614/*
615 * Transaction based IO.
616 * The file expects a single write which triggers the transaction, and then
617 * possibly a read which collects the result - which is stored in a
618 * file-local buffer.
619 */
Ingo Molnar76791ab2009-03-25 16:48:35 +0100620
621void simple_transaction_set(struct file *file, size_t n)
622{
623 struct simple_transaction_argresp *ar = file->private_data;
624
625 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
626
627 /*
628 * The barrier ensures that ar->size will really remain zero until
629 * ar->data is ready for reading.
630 */
631 smp_mb();
632 ar->size = n;
633}
634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
636{
637 struct simple_transaction_argresp *ar;
638 static DEFINE_SPINLOCK(simple_transaction_lock);
639
640 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
641 return ERR_PTR(-EFBIG);
642
643 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
644 if (!ar)
645 return ERR_PTR(-ENOMEM);
646
647 spin_lock(&simple_transaction_lock);
648
649 /* only one write allowed per open */
650 if (file->private_data) {
651 spin_unlock(&simple_transaction_lock);
652 free_page((unsigned long)ar);
653 return ERR_PTR(-EBUSY);
654 }
655
656 file->private_data = ar;
657
658 spin_unlock(&simple_transaction_lock);
659
660 if (copy_from_user(ar->data, buf, size))
661 return ERR_PTR(-EFAULT);
662
663 return ar->data;
664}
665
666ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
667{
668 struct simple_transaction_argresp *ar = file->private_data;
669
670 if (!ar)
671 return 0;
672 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
673}
674
675int simple_transaction_release(struct inode *inode, struct file *file)
676{
677 free_page((unsigned long)file->private_data);
678 return 0;
679}
680
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200681/* Simple attribute files */
682
683struct simple_attr {
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800684 int (*get)(void *, u64 *);
685 int (*set)(void *, u64);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200686 char get_buf[24]; /* enough to store a u64 and "\n\0" */
687 char set_buf[24];
688 void *data;
689 const char *fmt; /* format for read operation */
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800690 struct mutex mutex; /* protects access to these buffers */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200691};
692
693/* simple_attr_open is called by an actual attribute open file operation
694 * to set the attribute specific access operations. */
695int simple_attr_open(struct inode *inode, struct file *file,
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800696 int (*get)(void *, u64 *), int (*set)(void *, u64),
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200697 const char *fmt)
698{
699 struct simple_attr *attr;
700
701 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
702 if (!attr)
703 return -ENOMEM;
704
705 attr->get = get;
706 attr->set = set;
Theodore Ts'o8e18e292006-09-27 01:50:46 -0700707 attr->data = inode->i_private;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200708 attr->fmt = fmt;
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800709 mutex_init(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200710
711 file->private_data = attr;
712
713 return nonseekable_open(inode, file);
714}
715
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800716int simple_attr_release(struct inode *inode, struct file *file)
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200717{
718 kfree(file->private_data);
719 return 0;
720}
721
722/* read from the buffer that is filled with the get function */
723ssize_t simple_attr_read(struct file *file, char __user *buf,
724 size_t len, loff_t *ppos)
725{
726 struct simple_attr *attr;
727 size_t size;
728 ssize_t ret;
729
730 attr = file->private_data;
731
732 if (!attr->get)
733 return -EACCES;
734
Christoph Hellwig92613032008-02-08 04:20:27 -0800735 ret = mutex_lock_interruptible(&attr->mutex);
736 if (ret)
737 return ret;
738
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800739 if (*ppos) { /* continued read */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200740 size = strlen(attr->get_buf);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800741 } else { /* first read */
742 u64 val;
743 ret = attr->get(attr->data, &val);
744 if (ret)
745 goto out;
746
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200747 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800748 attr->fmt, (unsigned long long)val);
749 }
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200750
751 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800752out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800753 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200754 return ret;
755}
756
757/* interpret the buffer as a number to call the set function with */
758ssize_t simple_attr_write(struct file *file, const char __user *buf,
759 size_t len, loff_t *ppos)
760{
761 struct simple_attr *attr;
762 u64 val;
763 size_t size;
764 ssize_t ret;
765
766 attr = file->private_data;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200767 if (!attr->set)
768 return -EACCES;
769
Christoph Hellwig92613032008-02-08 04:20:27 -0800770 ret = mutex_lock_interruptible(&attr->mutex);
771 if (ret)
772 return ret;
773
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200774 ret = -EFAULT;
775 size = min(sizeof(attr->set_buf) - 1, len);
776 if (copy_from_user(attr->set_buf, buf, size))
777 goto out;
778
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200779 attr->set_buf[size] = '\0';
780 val = simple_strtol(attr->set_buf, NULL, 0);
Wu Fengguang05cc0ce2009-09-18 13:06:03 -0700781 ret = attr->set(attr->data, val);
782 if (ret == 0)
783 ret = len; /* on success, claim we got the whole input */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200784out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800785 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200786 return ret;
787}
788
Christoph Hellwig25961102007-10-21 16:42:05 -0700789/**
790 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
791 * @sb: filesystem to do the file handle conversion on
792 * @fid: file handle to convert
793 * @fh_len: length of the file handle in bytes
794 * @fh_type: type of file handle
795 * @get_inode: filesystem callback to retrieve inode
796 *
797 * This function decodes @fid as long as it has one of the well-known
798 * Linux filehandle types and calls @get_inode on it to retrieve the
799 * inode for the object specified in the file handle.
800 */
801struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
802 int fh_len, int fh_type, struct inode *(*get_inode)
803 (struct super_block *sb, u64 ino, u32 gen))
804{
805 struct inode *inode = NULL;
806
807 if (fh_len < 2)
808 return NULL;
809
810 switch (fh_type) {
811 case FILEID_INO32_GEN:
812 case FILEID_INO32_GEN_PARENT:
813 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
814 break;
815 }
816
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +0200817 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -0700818}
819EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
820
821/**
822 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
823 * @sb: filesystem to do the file handle conversion on
824 * @fid: file handle to convert
825 * @fh_len: length of the file handle in bytes
826 * @fh_type: type of file handle
827 * @get_inode: filesystem callback to retrieve inode
828 *
829 * This function decodes @fid as long as it has one of the well-known
830 * Linux filehandle types and calls @get_inode on it to retrieve the
831 * inode for the _parent_ object specified in the file handle if it
832 * is specified in the file handle, or NULL otherwise.
833 */
834struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
835 int fh_len, int fh_type, struct inode *(*get_inode)
836 (struct super_block *sb, u64 ino, u32 gen))
837{
838 struct inode *inode = NULL;
839
840 if (fh_len <= 2)
841 return NULL;
842
843 switch (fh_type) {
844 case FILEID_INO32_GEN_PARENT:
845 inode = get_inode(sb, fid->i32.parent_ino,
846 (fh_len > 3 ? fid->i32.parent_gen : 0));
847 break;
848 }
849
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +0200850 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -0700851}
852EXPORT_SYMBOL_GPL(generic_fh_to_parent);
853
Al Virod5aacad2009-06-07 14:56:44 -0400854int simple_fsync(struct file *file, struct dentry *dentry, int datasync)
855{
856 struct writeback_control wbc = {
857 .sync_mode = WB_SYNC_ALL,
858 .nr_to_write = 0, /* metadata-only; caller takes care of data */
859 };
860 struct inode *inode = dentry->d_inode;
861 int err;
862 int ret;
863
864 ret = sync_mapping_buffers(inode->i_mapping);
865 if (!(inode->i_state & I_DIRTY))
866 return ret;
867 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
868 return ret;
869
870 err = sync_inode(inode, &wbc);
871 if (ret == 0)
872 ret = err;
873 return ret;
874}
875EXPORT_SYMBOL(simple_fsync);
876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877EXPORT_SYMBOL(dcache_dir_close);
878EXPORT_SYMBOL(dcache_dir_lseek);
879EXPORT_SYMBOL(dcache_dir_open);
880EXPORT_SYMBOL(dcache_readdir);
881EXPORT_SYMBOL(generic_read_dir);
882EXPORT_SYMBOL(get_sb_pseudo);
Nick Pigginafddba42007-10-16 01:25:01 -0700883EXPORT_SYMBOL(simple_write_begin);
884EXPORT_SYMBOL(simple_write_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885EXPORT_SYMBOL(simple_dir_inode_operations);
886EXPORT_SYMBOL(simple_dir_operations);
887EXPORT_SYMBOL(simple_empty);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888EXPORT_SYMBOL(simple_fill_super);
889EXPORT_SYMBOL(simple_getattr);
890EXPORT_SYMBOL(simple_link);
891EXPORT_SYMBOL(simple_lookup);
892EXPORT_SYMBOL(simple_pin_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893EXPORT_SYMBOL(simple_readpage);
894EXPORT_SYMBOL(simple_release_fs);
895EXPORT_SYMBOL(simple_rename);
896EXPORT_SYMBOL(simple_rmdir);
897EXPORT_SYMBOL(simple_statfs);
898EXPORT_SYMBOL(simple_sync_file);
899EXPORT_SYMBOL(simple_unlink);
900EXPORT_SYMBOL(simple_read_from_buffer);
Jiri Slaby6a727b42010-05-01 23:51:22 +0200901EXPORT_SYMBOL(simple_write_to_buffer);
Akinobu Mita93b07112008-06-05 22:46:21 -0700902EXPORT_SYMBOL(memory_read_from_buffer);
Ingo Molnar76791ab2009-03-25 16:48:35 +0100903EXPORT_SYMBOL(simple_transaction_set);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904EXPORT_SYMBOL(simple_transaction_get);
905EXPORT_SYMBOL(simple_transaction_read);
906EXPORT_SYMBOL(simple_transaction_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200907EXPORT_SYMBOL_GPL(simple_attr_open);
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800908EXPORT_SYMBOL_GPL(simple_attr_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200909EXPORT_SYMBOL_GPL(simple_attr_read);
910EXPORT_SYMBOL_GPL(simple_attr_write);