blob: a3accdf528add9161bd6dccfb84f7c08ccb28437 [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>
npiggin@suse.de7bb46a62010-05-27 01:05:33 +100011#include <linux/quotaops.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080012#include <linux/mutex.h>
Christoph Hellwig25961102007-10-21 16:42:05 -070013#include <linux/exportfs.h>
Al Virod5aacad2009-06-07 14:56:44 -040014#include <linux/writeback.h>
15#include <linux/buffer_head.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080016
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#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
David Howells726c3342006-06-23 02:02:58 -070028int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070029{
David Howells726c3342006-06-23 02:02:58 -070030 buf->f_type = dentry->d_sb->s_magic;
Linus Torvalds1da177e2005-04-16 15:20:36 -070031 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{
Al Viro3ba13d12009-02-20 06:02:22 +000051 static const struct dentry_operations simple_dentry_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070052 .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
Linus Torvalds1da177e2005-04-16 15:20:36 -070062int dcache_dir_open(struct inode *inode, struct file *file)
63{
64 static struct qstr cursor_name = {.len = 1, .name = "."};
65
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080066 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
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{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080079 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 switch (origin) {
81 case 1:
82 offset += file->f_pos;
83 case 0:
84 if (offset >= 0)
85 break;
86 default:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080087 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 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);
Eric Dumazet5160ee62006-01-08 01:03:32 -080098 list_del(&cursor->d_u.d_child);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080099 p = file->f_path.dentry->d_subdirs.next;
100 while (n && p != &file->f_path.dentry->d_subdirs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 struct dentry *next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800102 next = list_entry(p, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 if (!d_unhashed(next) && next->d_inode)
104 n--;
105 p = p->next;
106 }
Eric Dumazet5160ee62006-01-08 01:03:32 -0800107 list_add_tail(&cursor->d_u.d_child, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 spin_unlock(&dcache_lock);
109 }
110 }
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800111 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 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{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800129 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 struct dentry *cursor = filp->private_data;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800131 struct list_head *p, *q = &cursor->d_u.d_child;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 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);
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700152 if (filp->f_pos == 2)
153 list_move(q, &dentry->d_subdirs);
154
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
156 struct dentry *next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800157 next = list_entry(p, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 if (d_unhashed(next) || !next->d_inode)
159 continue;
160
161 spin_unlock(&dcache_lock);
Ronni Nielsen0f8952c22007-05-09 06:44:57 +0200162 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)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 return 0;
167 spin_lock(&dcache_lock);
168 /* next is still alive */
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700169 list_move(q, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 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
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800183const struct file_operations simple_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 .open = dcache_dir_open,
185 .release = dcache_dir_close,
186 .llseek = dcache_dir_lseek,
187 .read = generic_read_dir,
188 .readdir = dcache_readdir,
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200189 .fsync = noop_fsync,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190};
191
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -0800192const struct inode_operations simple_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 .lookup = simple_lookup,
194};
195
Hugh Dickins759b9772007-03-05 00:30:28 -0800196static const struct super_operations simple_super_operations = {
197 .statfs = simple_statfs,
198};
199
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200/*
201 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
202 * will never be mountable)
203 */
Al Viro51139ad2010-07-25 23:47:46 +0400204struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
205 const struct super_operations *ops, unsigned long magic)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206{
207 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 struct dentry *dentry;
209 struct inode *root;
210 struct qstr d_name = {.name = name, .len = strlen(name)};
211
212 if (IS_ERR(s))
Al Viro51139ad2010-07-25 23:47:46 +0400213 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214
215 s->s_flags = MS_NOUSER;
Jeff Layton89a4eb42009-08-18 14:11:08 -0700216 s->s_maxbytes = MAX_LFS_FILESIZE;
Alex Nixon3971e1a2008-07-29 22:33:03 -0700217 s->s_blocksize = PAGE_SIZE;
218 s->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800220 s->s_op = ops ? ops : &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 s->s_time_gran = 1;
222 root = new_inode(s);
223 if (!root)
224 goto Enomem;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700225 /*
226 * since this is the first inode, make it number 1. New inodes created
227 * after this must take care not to collide with it (by passing
228 * max_reserved of 1 to iunique).
229 */
230 root->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
233 dentry = d_alloc(NULL, &d_name);
234 if (!dentry) {
235 iput(root);
236 goto Enomem;
237 }
238 dentry->d_sb = s;
239 dentry->d_parent = dentry;
240 d_instantiate(dentry, root);
241 s->s_root = dentry;
242 s->s_flags |= MS_ACTIVE;
Al Viro51139ad2010-07-25 23:47:46 +0400243 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
245Enomem:
Al Viro6f5bbff2009-05-06 01:34:22 -0400246 deactivate_locked_super(s);
Al Viro51139ad2010-07-25 23:47:46 +0400247 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248}
249
250int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
251{
252 struct inode *inode = old_dentry->d_inode;
253
254 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
Dave Hansend8c76e62006-09-30 23:29:04 -0700255 inc_nlink(inode);
Al Viro7de9c6ee2010-10-23 11:11:40 -0400256 ihold(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 dget(dentry);
258 d_instantiate(dentry, inode);
259 return 0;
260}
261
262static inline int simple_positive(struct dentry *dentry)
263{
264 return dentry->d_inode && !d_unhashed(dentry);
265}
266
267int simple_empty(struct dentry *dentry)
268{
269 struct dentry *child;
270 int ret = 0;
271
272 spin_lock(&dcache_lock);
Eric Dumazet5160ee62006-01-08 01:03:32 -0800273 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 if (simple_positive(child))
275 goto out;
276 ret = 1;
277out:
278 spin_unlock(&dcache_lock);
279 return ret;
280}
281
282int simple_unlink(struct inode *dir, struct dentry *dentry)
283{
284 struct inode *inode = dentry->d_inode;
285
286 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700287 drop_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 dput(dentry);
289 return 0;
290}
291
292int simple_rmdir(struct inode *dir, struct dentry *dentry)
293{
294 if (!simple_empty(dentry))
295 return -ENOTEMPTY;
296
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700297 drop_nlink(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 simple_unlink(dir, dentry);
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700299 drop_nlink(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 return 0;
301}
302
303int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
304 struct inode *new_dir, struct dentry *new_dentry)
305{
306 struct inode *inode = old_dentry->d_inode;
307 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
308
309 if (!simple_empty(new_dentry))
310 return -ENOTEMPTY;
311
312 if (new_dentry->d_inode) {
313 simple_unlink(new_dir, new_dentry);
314 if (they_are_dirs)
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700315 drop_nlink(old_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 } else if (they_are_dirs) {
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700317 drop_nlink(old_dir);
Dave Hansend8c76e62006-09-30 23:29:04 -0700318 inc_nlink(new_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 }
320
321 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
322 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
323
324 return 0;
325}
326
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000327/**
Christoph Hellwigeef23802010-06-04 11:30:01 +0200328 * simple_setattr - setattr for simple filesystem
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000329 * @dentry: dentry
330 * @iattr: iattr structure
331 *
332 * Returns 0 on success, -error on failure.
333 *
Christoph Hellwigeef23802010-06-04 11:30:01 +0200334 * simple_setattr is a simple ->setattr implementation without a proper
335 * implementation of size changes.
336 *
337 * It can either be used for in-memory filesystems or special files
338 * on simple regular filesystems. Anything that needs to change on-disk
339 * or wire state on size changes needs its own setattr method.
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000340 */
341int simple_setattr(struct dentry *dentry, struct iattr *iattr)
342{
343 struct inode *inode = dentry->d_inode;
344 int error;
345
Christoph Hellwigeef23802010-06-04 11:30:01 +0200346 WARN_ON_ONCE(inode->i_op->truncate);
347
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000348 error = inode_change_ok(inode, iattr);
349 if (error)
350 return error;
351
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200352 if (iattr->ia_valid & ATTR_SIZE)
353 truncate_setsize(inode, iattr->ia_size);
Christoph Hellwig6a1a90a2010-06-04 11:30:00 +0200354 setattr_copy(inode, iattr);
Christoph Hellwigeef23802010-06-04 11:30:01 +0200355 mark_inode_dirty(inode);
356 return 0;
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000357}
358EXPORT_SYMBOL(simple_setattr);
359
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360int simple_readpage(struct file *file, struct page *page)
361{
Pekka J Enbergc0d92cb2006-09-29 01:59:09 -0700362 clear_highpage(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 flush_dcache_page(page);
364 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 unlock_page(page);
366 return 0;
367}
368
Nick Pigginafddba42007-10-16 01:25:01 -0700369int simple_write_begin(struct file *file, struct address_space *mapping,
370 loff_t pos, unsigned len, unsigned flags,
371 struct page **pagep, void **fsdata)
372{
373 struct page *page;
374 pgoff_t index;
Nick Pigginafddba42007-10-16 01:25:01 -0700375
376 index = pos >> PAGE_CACHE_SHIFT;
Nick Pigginafddba42007-10-16 01:25:01 -0700377
Nick Piggin54566b22009-01-04 12:00:53 -0800378 page = grab_cache_page_write_begin(mapping, index, flags);
Nick Pigginafddba42007-10-16 01:25:01 -0700379 if (!page)
380 return -ENOMEM;
381
382 *pagep = page;
383
Boaz Harrosh193cf4b2010-01-12 16:18:08 +0200384 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
385 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
386
387 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
388 }
389 return 0;
Nick Pigginafddba42007-10-16 01:25:01 -0700390}
391
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200392/**
393 * simple_write_end - .write_end helper for non-block-device FSes
394 * @available: See .write_end of address_space_operations
395 * @file: "
396 * @mapping: "
397 * @pos: "
398 * @len: "
399 * @copied: "
400 * @page: "
401 * @fsdata: "
402 *
403 * simple_write_end does the minimum needed for updating a page after writing is
404 * done. It has the same API signature as the .write_end of
405 * address_space_operations vector. So it can just be set onto .write_end for
406 * FSes that don't need any other processing. i_mutex is assumed to be held.
407 * Block based filesystems should use generic_write_end().
408 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
409 * is not called, so a filesystem that actually does store data in .write_inode
410 * should extend on what's done here with a call to mark_inode_dirty() in the
411 * case that i_size has changed.
412 */
413int simple_write_end(struct file *file, struct address_space *mapping,
414 loff_t pos, unsigned len, unsigned copied,
415 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416{
417 struct inode *inode = page->mapping->host;
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200418 loff_t last_pos = pos + copied;
419
420 /* zero the stale part of the page if we did a short copy */
421 if (copied < len) {
422 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
423
424 zero_user(page, from + copied, len - copied);
425 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426
Nick Piggin955eff52007-02-20 13:58:08 -0800427 if (!PageUptodate(page))
428 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 /*
430 * No need to use i_size_read() here, the i_size
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800431 * cannot change under us because we hold the i_mutex.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 */
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200433 if (last_pos > inode->i_size)
434 i_size_write(inode, last_pos);
435
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 set_page_dirty(page);
Nick Pigginafddba42007-10-16 01:25:01 -0700437 unlock_page(page);
438 page_cache_release(page);
439
440 return copied;
441}
442
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700443/*
444 * the inodes created here are not hashed. If you use iunique to generate
445 * unique inode values later for this filesystem, then you must take care
446 * to pass it an appropriate max_reserved value to avoid collisions.
447 */
Roberto Sassu7d683a02010-06-03 11:58:28 +0200448int simple_fill_super(struct super_block *s, unsigned long magic,
449 struct tree_descr *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 struct inode *inode;
452 struct dentry *root;
453 struct dentry *dentry;
454 int i;
455
456 s->s_blocksize = PAGE_CACHE_SIZE;
457 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
458 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800459 s->s_op = &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 s->s_time_gran = 1;
461
462 inode = new_inode(s);
463 if (!inode)
464 return -ENOMEM;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700465 /*
466 * because the root inode is 1, the files array must not contain an
467 * entry at index 1
468 */
469 inode->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 inode->i_mode = S_IFDIR | 0755;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
472 inode->i_op = &simple_dir_inode_operations;
473 inode->i_fop = &simple_dir_operations;
Vincent Hanquez7656f322006-02-03 03:04:48 -0800474 inode->i_nlink = 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 root = d_alloc_root(inode);
476 if (!root) {
477 iput(inode);
478 return -ENOMEM;
479 }
480 for (i = 0; !files->name || files->name[0]; i++, files++) {
481 if (!files->name)
482 continue;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700483
484 /* warn if it tries to conflict with the root inode */
485 if (unlikely(i == 1))
486 printk(KERN_WARNING "%s: %s passed in a files array"
487 "with an index of 1!\n", __func__,
488 s->s_type->name);
489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 dentry = d_alloc_name(root, files->name);
491 if (!dentry)
492 goto out;
493 inode = new_inode(s);
494 if (!inode)
495 goto out;
496 inode->i_mode = S_IFREG | files->mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
498 inode->i_fop = files->ops;
499 inode->i_ino = i;
500 d_add(dentry, inode);
501 }
502 s->s_root = root;
503 return 0;
504out:
505 d_genocide(root);
506 dput(root);
507 return -ENOMEM;
508}
509
510static DEFINE_SPINLOCK(pin_fs_lock);
511
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400512int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
514 struct vfsmount *mnt = NULL;
515 spin_lock(&pin_fs_lock);
516 if (unlikely(!*mount)) {
517 spin_unlock(&pin_fs_lock);
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400518 mnt = vfs_kern_mount(type, 0, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 if (IS_ERR(mnt))
520 return PTR_ERR(mnt);
521 spin_lock(&pin_fs_lock);
522 if (!*mount)
523 *mount = mnt;
524 }
525 mntget(*mount);
526 ++*count;
527 spin_unlock(&pin_fs_lock);
528 mntput(mnt);
529 return 0;
530}
531
532void simple_release_fs(struct vfsmount **mount, int *count)
533{
534 struct vfsmount *mnt;
535 spin_lock(&pin_fs_lock);
536 mnt = *mount;
537 if (!--*count)
538 *mount = NULL;
539 spin_unlock(&pin_fs_lock);
540 mntput(mnt);
541}
542
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700543/**
544 * simple_read_from_buffer - copy data from the buffer to user space
545 * @to: the user space buffer to read to
546 * @count: the maximum number of bytes to read
547 * @ppos: the current position in the buffer
548 * @from: the buffer to read from
549 * @available: the size of the buffer
550 *
551 * The simple_read_from_buffer() function reads up to @count bytes from the
552 * buffer @from at offset @ppos into the user space address starting at @to.
553 *
554 * On success, the number of bytes read is returned and the offset @ppos is
555 * advanced by this number, or negative value is returned on error.
556 **/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
558 const void *from, size_t available)
559{
560 loff_t pos = *ppos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700561 size_t ret;
562
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 if (pos < 0)
564 return -EINVAL;
Steven Rostedt14be2742009-09-18 13:05:42 -0700565 if (pos >= available || !count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 return 0;
567 if (count > available - pos)
568 count = available - pos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700569 ret = copy_to_user(to, from + pos, count);
570 if (ret == count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 return -EFAULT;
Steven Rostedt14be2742009-09-18 13:05:42 -0700572 count -= ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 *ppos = pos + count;
574 return count;
575}
576
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700577/**
Jiri Slaby6a727b42010-05-01 23:51:22 +0200578 * simple_write_to_buffer - copy data from user space to the buffer
579 * @to: the buffer to write to
580 * @available: the size of the buffer
581 * @ppos: the current position in the buffer
582 * @from: the user space buffer to read from
583 * @count: the maximum number of bytes to read
584 *
585 * The simple_write_to_buffer() function reads up to @count bytes from the user
586 * space address starting at @from into the buffer @to at offset @ppos.
587 *
588 * On success, the number of bytes written is returned and the offset @ppos is
589 * advanced by this number, or negative value is returned on error.
590 **/
591ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
592 const void __user *from, size_t count)
593{
594 loff_t pos = *ppos;
595 size_t res;
596
597 if (pos < 0)
598 return -EINVAL;
599 if (pos >= available || !count)
600 return 0;
601 if (count > available - pos)
602 count = available - pos;
603 res = copy_from_user(to + pos, from, count);
604 if (res == count)
605 return -EFAULT;
606 count -= res;
607 *ppos = pos + count;
608 return count;
609}
610
611/**
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700612 * memory_read_from_buffer - copy data from the buffer
613 * @to: the kernel space buffer to read to
614 * @count: the maximum number of bytes to read
615 * @ppos: the current position in the buffer
616 * @from: the buffer to read from
617 * @available: the size of the buffer
618 *
619 * The memory_read_from_buffer() function reads up to @count bytes from the
620 * buffer @from at offset @ppos into the kernel space address starting at @to.
621 *
622 * On success, the number of bytes read is returned and the offset @ppos is
623 * advanced by this number, or negative value is returned on error.
624 **/
Akinobu Mita93b07112008-06-05 22:46:21 -0700625ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
626 const void *from, size_t available)
627{
628 loff_t pos = *ppos;
629
630 if (pos < 0)
631 return -EINVAL;
632 if (pos >= available)
633 return 0;
634 if (count > available - pos)
635 count = available - pos;
636 memcpy(to, from + pos, count);
637 *ppos = pos + count;
638
639 return count;
640}
641
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642/*
643 * Transaction based IO.
644 * The file expects a single write which triggers the transaction, and then
645 * possibly a read which collects the result - which is stored in a
646 * file-local buffer.
647 */
Ingo Molnar76791ab2009-03-25 16:48:35 +0100648
649void simple_transaction_set(struct file *file, size_t n)
650{
651 struct simple_transaction_argresp *ar = file->private_data;
652
653 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
654
655 /*
656 * The barrier ensures that ar->size will really remain zero until
657 * ar->data is ready for reading.
658 */
659 smp_mb();
660 ar->size = n;
661}
662
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
664{
665 struct simple_transaction_argresp *ar;
666 static DEFINE_SPINLOCK(simple_transaction_lock);
667
668 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
669 return ERR_PTR(-EFBIG);
670
671 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
672 if (!ar)
673 return ERR_PTR(-ENOMEM);
674
675 spin_lock(&simple_transaction_lock);
676
677 /* only one write allowed per open */
678 if (file->private_data) {
679 spin_unlock(&simple_transaction_lock);
680 free_page((unsigned long)ar);
681 return ERR_PTR(-EBUSY);
682 }
683
684 file->private_data = ar;
685
686 spin_unlock(&simple_transaction_lock);
687
688 if (copy_from_user(ar->data, buf, size))
689 return ERR_PTR(-EFAULT);
690
691 return ar->data;
692}
693
694ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
695{
696 struct simple_transaction_argresp *ar = file->private_data;
697
698 if (!ar)
699 return 0;
700 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
701}
702
703int simple_transaction_release(struct inode *inode, struct file *file)
704{
705 free_page((unsigned long)file->private_data);
706 return 0;
707}
708
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200709/* Simple attribute files */
710
711struct simple_attr {
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800712 int (*get)(void *, u64 *);
713 int (*set)(void *, u64);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200714 char get_buf[24]; /* enough to store a u64 and "\n\0" */
715 char set_buf[24];
716 void *data;
717 const char *fmt; /* format for read operation */
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800718 struct mutex mutex; /* protects access to these buffers */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200719};
720
721/* simple_attr_open is called by an actual attribute open file operation
722 * to set the attribute specific access operations. */
723int simple_attr_open(struct inode *inode, struct file *file,
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800724 int (*get)(void *, u64 *), int (*set)(void *, u64),
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200725 const char *fmt)
726{
727 struct simple_attr *attr;
728
729 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
730 if (!attr)
731 return -ENOMEM;
732
733 attr->get = get;
734 attr->set = set;
Theodore Ts'o8e18e292006-09-27 01:50:46 -0700735 attr->data = inode->i_private;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200736 attr->fmt = fmt;
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800737 mutex_init(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200738
739 file->private_data = attr;
740
741 return nonseekable_open(inode, file);
742}
743
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800744int simple_attr_release(struct inode *inode, struct file *file)
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200745{
746 kfree(file->private_data);
747 return 0;
748}
749
750/* read from the buffer that is filled with the get function */
751ssize_t simple_attr_read(struct file *file, char __user *buf,
752 size_t len, loff_t *ppos)
753{
754 struct simple_attr *attr;
755 size_t size;
756 ssize_t ret;
757
758 attr = file->private_data;
759
760 if (!attr->get)
761 return -EACCES;
762
Christoph Hellwig92613032008-02-08 04:20:27 -0800763 ret = mutex_lock_interruptible(&attr->mutex);
764 if (ret)
765 return ret;
766
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800767 if (*ppos) { /* continued read */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200768 size = strlen(attr->get_buf);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800769 } else { /* first read */
770 u64 val;
771 ret = attr->get(attr->data, &val);
772 if (ret)
773 goto out;
774
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200775 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800776 attr->fmt, (unsigned long long)val);
777 }
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200778
779 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800780out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800781 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200782 return ret;
783}
784
785/* interpret the buffer as a number to call the set function with */
786ssize_t simple_attr_write(struct file *file, const char __user *buf,
787 size_t len, loff_t *ppos)
788{
789 struct simple_attr *attr;
790 u64 val;
791 size_t size;
792 ssize_t ret;
793
794 attr = file->private_data;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200795 if (!attr->set)
796 return -EACCES;
797
Christoph Hellwig92613032008-02-08 04:20:27 -0800798 ret = mutex_lock_interruptible(&attr->mutex);
799 if (ret)
800 return ret;
801
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200802 ret = -EFAULT;
803 size = min(sizeof(attr->set_buf) - 1, len);
804 if (copy_from_user(attr->set_buf, buf, size))
805 goto out;
806
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200807 attr->set_buf[size] = '\0';
808 val = simple_strtol(attr->set_buf, NULL, 0);
Wu Fengguang05cc0ce2009-09-18 13:06:03 -0700809 ret = attr->set(attr->data, val);
810 if (ret == 0)
811 ret = len; /* on success, claim we got the whole input */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200812out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800813 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200814 return ret;
815}
816
Christoph Hellwig25961102007-10-21 16:42:05 -0700817/**
818 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
819 * @sb: filesystem to do the file handle conversion on
820 * @fid: file handle to convert
821 * @fh_len: length of the file handle in bytes
822 * @fh_type: type of file handle
823 * @get_inode: filesystem callback to retrieve inode
824 *
825 * This function decodes @fid as long as it has one of the well-known
826 * Linux filehandle types and calls @get_inode on it to retrieve the
827 * inode for the object specified in the file handle.
828 */
829struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
830 int fh_len, int fh_type, struct inode *(*get_inode)
831 (struct super_block *sb, u64 ino, u32 gen))
832{
833 struct inode *inode = NULL;
834
835 if (fh_len < 2)
836 return NULL;
837
838 switch (fh_type) {
839 case FILEID_INO32_GEN:
840 case FILEID_INO32_GEN_PARENT:
841 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
842 break;
843 }
844
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +0200845 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -0700846}
847EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
848
849/**
850 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
851 * @sb: filesystem to do the file handle conversion on
852 * @fid: file handle to convert
853 * @fh_len: length of the file handle in bytes
854 * @fh_type: type of file handle
855 * @get_inode: filesystem callback to retrieve inode
856 *
857 * This function decodes @fid as long as it has one of the well-known
858 * Linux filehandle types and calls @get_inode on it to retrieve the
859 * inode for the _parent_ object specified in the file handle if it
860 * is specified in the file handle, or NULL otherwise.
861 */
862struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
863 int fh_len, int fh_type, struct inode *(*get_inode)
864 (struct super_block *sb, u64 ino, u32 gen))
865{
866 struct inode *inode = NULL;
867
868 if (fh_len <= 2)
869 return NULL;
870
871 switch (fh_type) {
872 case FILEID_INO32_GEN_PARENT:
873 inode = get_inode(sb, fid->i32.parent_ino,
874 (fh_len > 3 ? fid->i32.parent_gen : 0));
875 break;
876 }
877
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +0200878 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -0700879}
880EXPORT_SYMBOL_GPL(generic_fh_to_parent);
881
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200882/**
883 * generic_file_fsync - generic fsync implementation for simple filesystems
884 * @file: file to synchronize
885 * @datasync: only synchronize essential metadata if true
886 *
887 * This is a generic implementation of the fsync method for simple
888 * filesystems which track all non-inode metadata in the buffers list
889 * hanging off the address_space structure.
890 */
891int generic_file_fsync(struct file *file, int datasync)
Al Virod5aacad2009-06-07 14:56:44 -0400892{
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200893 struct inode *inode = file->f_mapping->host;
Al Virod5aacad2009-06-07 14:56:44 -0400894 int err;
895 int ret;
896
897 ret = sync_mapping_buffers(inode->i_mapping);
898 if (!(inode->i_state & I_DIRTY))
899 return ret;
900 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
901 return ret;
902
Christoph Hellwigc37650162010-10-06 10:48:20 +0200903 err = sync_inode_metadata(inode, 1);
Al Virod5aacad2009-06-07 14:56:44 -0400904 if (ret == 0)
905 ret = err;
906 return ret;
907}
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200908EXPORT_SYMBOL(generic_file_fsync);
909
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -0700910/**
911 * generic_check_addressable - Check addressability of file system
912 * @blocksize_bits: log of file system block size
913 * @num_blocks: number of blocks in file system
914 *
915 * Determine whether a file system with @num_blocks blocks (and a
916 * block size of 2**@blocksize_bits) is addressable by the sector_t
917 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
918 */
919int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
920{
921 u64 last_fs_block = num_blocks - 1;
Joel Beckera33f13e2010-08-16 12:10:17 -0700922 u64 last_fs_page =
923 last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -0700924
925 if (unlikely(num_blocks == 0))
926 return 0;
927
928 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
929 return -EINVAL;
930
Joel Beckera33f13e2010-08-16 12:10:17 -0700931 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
932 (last_fs_page > (pgoff_t)(~0ULL))) {
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -0700933 return -EFBIG;
934 }
935 return 0;
936}
937EXPORT_SYMBOL(generic_check_addressable);
938
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200939/*
940 * No-op implementation of ->fsync for in-memory filesystems.
941 */
942int noop_fsync(struct file *file, int datasync)
943{
944 return 0;
945}
Al Virod5aacad2009-06-07 14:56:44 -0400946
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947EXPORT_SYMBOL(dcache_dir_close);
948EXPORT_SYMBOL(dcache_dir_lseek);
949EXPORT_SYMBOL(dcache_dir_open);
950EXPORT_SYMBOL(dcache_readdir);
951EXPORT_SYMBOL(generic_read_dir);
Al Viro51139ad2010-07-25 23:47:46 +0400952EXPORT_SYMBOL(mount_pseudo);
Nick Pigginafddba42007-10-16 01:25:01 -0700953EXPORT_SYMBOL(simple_write_begin);
954EXPORT_SYMBOL(simple_write_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955EXPORT_SYMBOL(simple_dir_inode_operations);
956EXPORT_SYMBOL(simple_dir_operations);
957EXPORT_SYMBOL(simple_empty);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958EXPORT_SYMBOL(simple_fill_super);
959EXPORT_SYMBOL(simple_getattr);
960EXPORT_SYMBOL(simple_link);
961EXPORT_SYMBOL(simple_lookup);
962EXPORT_SYMBOL(simple_pin_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963EXPORT_SYMBOL(simple_readpage);
964EXPORT_SYMBOL(simple_release_fs);
965EXPORT_SYMBOL(simple_rename);
966EXPORT_SYMBOL(simple_rmdir);
967EXPORT_SYMBOL(simple_statfs);
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200968EXPORT_SYMBOL(noop_fsync);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969EXPORT_SYMBOL(simple_unlink);
970EXPORT_SYMBOL(simple_read_from_buffer);
Jiri Slaby6a727b42010-05-01 23:51:22 +0200971EXPORT_SYMBOL(simple_write_to_buffer);
Akinobu Mita93b07112008-06-05 22:46:21 -0700972EXPORT_SYMBOL(memory_read_from_buffer);
Ingo Molnar76791ab2009-03-25 16:48:35 +0100973EXPORT_SYMBOL(simple_transaction_set);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974EXPORT_SYMBOL(simple_transaction_get);
975EXPORT_SYMBOL(simple_transaction_read);
976EXPORT_SYMBOL(simple_transaction_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200977EXPORT_SYMBOL_GPL(simple_attr_open);
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800978EXPORT_SYMBOL_GPL(simple_attr_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200979EXPORT_SYMBOL_GPL(simple_attr_read);
980EXPORT_SYMBOL_GPL(simple_attr_write);