blob: 433e7139c23aa179905fb3e35636f3c2600098d4 [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
Nick Pigginda502952011-01-07 17:49:33 +110019static inline int simple_positive(struct dentry *dentry)
20{
21 return dentry->d_inode && !d_unhashed(dentry);
22}
23
Linus Torvalds1da177e2005-04-16 15:20:36 -070024int 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
David Howells726c3342006-06-23 02:02:58 -070033int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070034{
David Howells726c3342006-06-23 02:02:58 -070035 buf->f_type = dentry->d_sb->s_magic;
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 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 */
Nick Pigginfe15ce42011-01-07 17:49:23 +110045static int simple_delete_dentry(const struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -070046{
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{
Al Viro3ba13d12009-02-20 06:02:22 +000056 static const struct dentry_operations simple_dentry_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 .d_delete = simple_delete_dentry,
58 };
59
60 if (dentry->d_name.len > NAME_MAX)
61 return ERR_PTR(-ENAMETOOLONG);
62 dentry->d_op = &simple_dentry_operations;
63 d_add(dentry, NULL);
64 return NULL;
65}
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067int dcache_dir_open(struct inode *inode, struct file *file)
68{
69 static struct qstr cursor_name = {.len = 1, .name = "."};
70
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080071 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
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{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080084 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085 switch (origin) {
86 case 1:
87 offset += file->f_pos;
88 case 0:
89 if (offset >= 0)
90 break;
91 default:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080092 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 return -EINVAL;
94 }
95 if (offset != file->f_pos) {
96 file->f_pos = offset;
97 if (file->f_pos >= 2) {
98 struct list_head *p;
99 struct dentry *cursor = file->private_data;
100 loff_t n = file->f_pos - 2;
101
102 spin_lock(&dcache_lock);
Eric Dumazet5160ee62006-01-08 01:03:32 -0800103 list_del(&cursor->d_u.d_child);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800104 p = file->f_path.dentry->d_subdirs.next;
105 while (n && p != &file->f_path.dentry->d_subdirs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 struct dentry *next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800107 next = list_entry(p, struct dentry, d_u.d_child);
Nick Pigginda502952011-01-07 17:49:33 +1100108 spin_lock(&next->d_lock);
109 if (simple_positive(next))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110 n--;
Nick Pigginda502952011-01-07 17:49:33 +1100111 spin_unlock(&next->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 p = p->next;
113 }
Eric Dumazet5160ee62006-01-08 01:03:32 -0800114 list_add_tail(&cursor->d_u.d_child, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 spin_unlock(&dcache_lock);
116 }
117 }
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800118 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 return offset;
120}
121
122/* Relationship between i_mode and the DT_xxx types */
123static inline unsigned char dt_type(struct inode *inode)
124{
125 return (inode->i_mode >> 12) & 15;
126}
127
128/*
129 * Directory is locked and all positive dentries in it are safe, since
130 * for ramfs-type trees they can't go away without unlink() or rmdir(),
131 * both impossible due to the lock on directory.
132 */
133
134int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
135{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800136 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 struct dentry *cursor = filp->private_data;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800138 struct list_head *p, *q = &cursor->d_u.d_child;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 ino_t ino;
140 int i = filp->f_pos;
141
142 switch (i) {
143 case 0:
144 ino = dentry->d_inode->i_ino;
145 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
146 break;
147 filp->f_pos++;
148 i++;
149 /* fallthrough */
150 case 1:
151 ino = parent_ino(dentry);
152 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
153 break;
154 filp->f_pos++;
155 i++;
156 /* fallthrough */
157 default:
158 spin_lock(&dcache_lock);
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700159 if (filp->f_pos == 2)
160 list_move(q, &dentry->d_subdirs);
161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
163 struct dentry *next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800164 next = list_entry(p, struct dentry, d_u.d_child);
Nick Pigginda502952011-01-07 17:49:33 +1100165 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
166 if (!simple_positive(next)) {
167 spin_unlock(&next->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 continue;
Nick Pigginda502952011-01-07 17:49:33 +1100169 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
Nick Pigginda502952011-01-07 17:49:33 +1100171 spin_unlock(&next->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 spin_unlock(&dcache_lock);
Ronni Nielsen0f8952c22007-05-09 06:44:57 +0200173 if (filldir(dirent, next->d_name.name,
174 next->d_name.len, filp->f_pos,
175 next->d_inode->i_ino,
176 dt_type(next->d_inode)) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 return 0;
178 spin_lock(&dcache_lock);
179 /* next is still alive */
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700180 list_move(q, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 p = q;
182 filp->f_pos++;
183 }
184 spin_unlock(&dcache_lock);
185 }
186 return 0;
187}
188
189ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
190{
191 return -EISDIR;
192}
193
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800194const struct file_operations simple_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 .open = dcache_dir_open,
196 .release = dcache_dir_close,
197 .llseek = dcache_dir_lseek,
198 .read = generic_read_dir,
199 .readdir = dcache_readdir,
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200200 .fsync = noop_fsync,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201};
202
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -0800203const struct inode_operations simple_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 .lookup = simple_lookup,
205};
206
Hugh Dickins759b9772007-03-05 00:30:28 -0800207static const struct super_operations simple_super_operations = {
208 .statfs = simple_statfs,
209};
210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211/*
212 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
213 * will never be mountable)
214 */
Al Viro51139ad2010-07-25 23:47:46 +0400215struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
216 const struct super_operations *ops, unsigned long magic)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217{
218 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 struct dentry *dentry;
220 struct inode *root;
221 struct qstr d_name = {.name = name, .len = strlen(name)};
222
223 if (IS_ERR(s))
Al Viro51139ad2010-07-25 23:47:46 +0400224 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
226 s->s_flags = MS_NOUSER;
Jeff Layton89a4eb42009-08-18 14:11:08 -0700227 s->s_maxbytes = MAX_LFS_FILESIZE;
Alex Nixon3971e1a2008-07-29 22:33:03 -0700228 s->s_blocksize = PAGE_SIZE;
229 s->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800231 s->s_op = ops ? ops : &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 s->s_time_gran = 1;
233 root = new_inode(s);
234 if (!root)
235 goto Enomem;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700236 /*
237 * since this is the first inode, make it number 1. New inodes created
238 * after this must take care not to collide with it (by passing
239 * max_reserved of 1 to iunique).
240 */
241 root->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
244 dentry = d_alloc(NULL, &d_name);
245 if (!dentry) {
246 iput(root);
247 goto Enomem;
248 }
249 dentry->d_sb = s;
250 dentry->d_parent = dentry;
251 d_instantiate(dentry, root);
252 s->s_root = dentry;
253 s->s_flags |= MS_ACTIVE;
Al Viro51139ad2010-07-25 23:47:46 +0400254 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256Enomem:
Al Viro6f5bbff2009-05-06 01:34:22 -0400257 deactivate_locked_super(s);
Al Viro51139ad2010-07-25 23:47:46 +0400258 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259}
260
261int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
262{
263 struct inode *inode = old_dentry->d_inode;
264
265 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
Dave Hansend8c76e62006-09-30 23:29:04 -0700266 inc_nlink(inode);
Al Viro7de9c6ee2010-10-23 11:11:40 -0400267 ihold(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 dget(dentry);
269 d_instantiate(dentry, inode);
270 return 0;
271}
272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273int simple_empty(struct dentry *dentry)
274{
275 struct dentry *child;
276 int ret = 0;
277
278 spin_lock(&dcache_lock);
Nick Pigginda502952011-01-07 17:49:33 +1100279 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
280 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
281 if (simple_positive(child)) {
282 spin_unlock(&child->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 goto out;
Nick Pigginda502952011-01-07 17:49:33 +1100284 }
285 spin_unlock(&child->d_lock);
286 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 ret = 1;
288out:
289 spin_unlock(&dcache_lock);
290 return ret;
291}
292
293int simple_unlink(struct inode *dir, struct dentry *dentry)
294{
295 struct inode *inode = dentry->d_inode;
296
297 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700298 drop_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 dput(dentry);
300 return 0;
301}
302
303int simple_rmdir(struct inode *dir, struct dentry *dentry)
304{
305 if (!simple_empty(dentry))
306 return -ENOTEMPTY;
307
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700308 drop_nlink(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 simple_unlink(dir, dentry);
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700310 drop_nlink(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 return 0;
312}
313
314int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
315 struct inode *new_dir, struct dentry *new_dentry)
316{
317 struct inode *inode = old_dentry->d_inode;
318 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
319
320 if (!simple_empty(new_dentry))
321 return -ENOTEMPTY;
322
323 if (new_dentry->d_inode) {
324 simple_unlink(new_dir, new_dentry);
325 if (they_are_dirs)
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700326 drop_nlink(old_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 } else if (they_are_dirs) {
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700328 drop_nlink(old_dir);
Dave Hansend8c76e62006-09-30 23:29:04 -0700329 inc_nlink(new_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 }
331
332 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
333 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
334
335 return 0;
336}
337
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000338/**
Christoph Hellwigeef23802010-06-04 11:30:01 +0200339 * simple_setattr - setattr for simple filesystem
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000340 * @dentry: dentry
341 * @iattr: iattr structure
342 *
343 * Returns 0 on success, -error on failure.
344 *
Christoph Hellwigeef23802010-06-04 11:30:01 +0200345 * simple_setattr is a simple ->setattr implementation without a proper
346 * implementation of size changes.
347 *
348 * It can either be used for in-memory filesystems or special files
349 * on simple regular filesystems. Anything that needs to change on-disk
350 * or wire state on size changes needs its own setattr method.
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000351 */
352int simple_setattr(struct dentry *dentry, struct iattr *iattr)
353{
354 struct inode *inode = dentry->d_inode;
355 int error;
356
Christoph Hellwigeef23802010-06-04 11:30:01 +0200357 WARN_ON_ONCE(inode->i_op->truncate);
358
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000359 error = inode_change_ok(inode, iattr);
360 if (error)
361 return error;
362
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200363 if (iattr->ia_valid & ATTR_SIZE)
364 truncate_setsize(inode, iattr->ia_size);
Christoph Hellwig6a1a90a2010-06-04 11:30:00 +0200365 setattr_copy(inode, iattr);
Christoph Hellwigeef23802010-06-04 11:30:01 +0200366 mark_inode_dirty(inode);
367 return 0;
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000368}
369EXPORT_SYMBOL(simple_setattr);
370
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371int simple_readpage(struct file *file, struct page *page)
372{
Pekka J Enbergc0d92cb2006-09-29 01:59:09 -0700373 clear_highpage(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 flush_dcache_page(page);
375 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 unlock_page(page);
377 return 0;
378}
379
Nick Pigginafddba42007-10-16 01:25:01 -0700380int simple_write_begin(struct file *file, struct address_space *mapping,
381 loff_t pos, unsigned len, unsigned flags,
382 struct page **pagep, void **fsdata)
383{
384 struct page *page;
385 pgoff_t index;
Nick Pigginafddba42007-10-16 01:25:01 -0700386
387 index = pos >> PAGE_CACHE_SHIFT;
Nick Pigginafddba42007-10-16 01:25:01 -0700388
Nick Piggin54566b22009-01-04 12:00:53 -0800389 page = grab_cache_page_write_begin(mapping, index, flags);
Nick Pigginafddba42007-10-16 01:25:01 -0700390 if (!page)
391 return -ENOMEM;
392
393 *pagep = page;
394
Boaz Harrosh193cf4b2010-01-12 16:18:08 +0200395 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
396 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
397
398 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
399 }
400 return 0;
Nick Pigginafddba42007-10-16 01:25:01 -0700401}
402
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200403/**
404 * simple_write_end - .write_end helper for non-block-device FSes
405 * @available: See .write_end of address_space_operations
406 * @file: "
407 * @mapping: "
408 * @pos: "
409 * @len: "
410 * @copied: "
411 * @page: "
412 * @fsdata: "
413 *
414 * simple_write_end does the minimum needed for updating a page after writing is
415 * done. It has the same API signature as the .write_end of
416 * address_space_operations vector. So it can just be set onto .write_end for
417 * FSes that don't need any other processing. i_mutex is assumed to be held.
418 * Block based filesystems should use generic_write_end().
419 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
420 * is not called, so a filesystem that actually does store data in .write_inode
421 * should extend on what's done here with a call to mark_inode_dirty() in the
422 * case that i_size has changed.
423 */
424int simple_write_end(struct file *file, struct address_space *mapping,
425 loff_t pos, unsigned len, unsigned copied,
426 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427{
428 struct inode *inode = page->mapping->host;
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200429 loff_t last_pos = pos + copied;
430
431 /* zero the stale part of the page if we did a short copy */
432 if (copied < len) {
433 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
434
435 zero_user(page, from + copied, len - copied);
436 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
Nick Piggin955eff52007-02-20 13:58:08 -0800438 if (!PageUptodate(page))
439 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 /*
441 * No need to use i_size_read() here, the i_size
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800442 * cannot change under us because we hold the i_mutex.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 */
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200444 if (last_pos > inode->i_size)
445 i_size_write(inode, last_pos);
446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 set_page_dirty(page);
Nick Pigginafddba42007-10-16 01:25:01 -0700448 unlock_page(page);
449 page_cache_release(page);
450
451 return copied;
452}
453
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700454/*
455 * the inodes created here are not hashed. If you use iunique to generate
456 * unique inode values later for this filesystem, then you must take care
457 * to pass it an appropriate max_reserved value to avoid collisions.
458 */
Roberto Sassu7d683a02010-06-03 11:58:28 +0200459int simple_fill_super(struct super_block *s, unsigned long magic,
460 struct tree_descr *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 struct inode *inode;
463 struct dentry *root;
464 struct dentry *dentry;
465 int i;
466
467 s->s_blocksize = PAGE_CACHE_SIZE;
468 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
469 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800470 s->s_op = &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 s->s_time_gran = 1;
472
473 inode = new_inode(s);
474 if (!inode)
475 return -ENOMEM;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700476 /*
477 * because the root inode is 1, the files array must not contain an
478 * entry at index 1
479 */
480 inode->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 inode->i_mode = S_IFDIR | 0755;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
483 inode->i_op = &simple_dir_inode_operations;
484 inode->i_fop = &simple_dir_operations;
Vincent Hanquez7656f322006-02-03 03:04:48 -0800485 inode->i_nlink = 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 root = d_alloc_root(inode);
487 if (!root) {
488 iput(inode);
489 return -ENOMEM;
490 }
491 for (i = 0; !files->name || files->name[0]; i++, files++) {
492 if (!files->name)
493 continue;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700494
495 /* warn if it tries to conflict with the root inode */
496 if (unlikely(i == 1))
497 printk(KERN_WARNING "%s: %s passed in a files array"
498 "with an index of 1!\n", __func__,
499 s->s_type->name);
500
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 dentry = d_alloc_name(root, files->name);
502 if (!dentry)
503 goto out;
504 inode = new_inode(s);
505 if (!inode)
506 goto out;
507 inode->i_mode = S_IFREG | files->mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
509 inode->i_fop = files->ops;
510 inode->i_ino = i;
511 d_add(dentry, inode);
512 }
513 s->s_root = root;
514 return 0;
515out:
516 d_genocide(root);
517 dput(root);
518 return -ENOMEM;
519}
520
521static DEFINE_SPINLOCK(pin_fs_lock);
522
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400523int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524{
525 struct vfsmount *mnt = NULL;
526 spin_lock(&pin_fs_lock);
527 if (unlikely(!*mount)) {
528 spin_unlock(&pin_fs_lock);
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400529 mnt = vfs_kern_mount(type, 0, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 if (IS_ERR(mnt))
531 return PTR_ERR(mnt);
532 spin_lock(&pin_fs_lock);
533 if (!*mount)
534 *mount = mnt;
535 }
536 mntget(*mount);
537 ++*count;
538 spin_unlock(&pin_fs_lock);
539 mntput(mnt);
540 return 0;
541}
542
543void simple_release_fs(struct vfsmount **mount, int *count)
544{
545 struct vfsmount *mnt;
546 spin_lock(&pin_fs_lock);
547 mnt = *mount;
548 if (!--*count)
549 *mount = NULL;
550 spin_unlock(&pin_fs_lock);
551 mntput(mnt);
552}
553
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700554/**
555 * simple_read_from_buffer - copy data from the buffer to user space
556 * @to: the user space buffer to read to
557 * @count: the maximum number of bytes to read
558 * @ppos: the current position in the buffer
559 * @from: the buffer to read from
560 * @available: the size of the buffer
561 *
562 * The simple_read_from_buffer() function reads up to @count bytes from the
563 * buffer @from at offset @ppos into the user space address starting at @to.
564 *
565 * On success, the number of bytes read is returned and the offset @ppos is
566 * advanced by this number, or negative value is returned on error.
567 **/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
569 const void *from, size_t available)
570{
571 loff_t pos = *ppos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700572 size_t ret;
573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 if (pos < 0)
575 return -EINVAL;
Steven Rostedt14be2742009-09-18 13:05:42 -0700576 if (pos >= available || !count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 return 0;
578 if (count > available - pos)
579 count = available - pos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700580 ret = copy_to_user(to, from + pos, count);
581 if (ret == count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 return -EFAULT;
Steven Rostedt14be2742009-09-18 13:05:42 -0700583 count -= ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 *ppos = pos + count;
585 return count;
586}
587
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700588/**
Jiri Slaby6a727b42010-05-01 23:51:22 +0200589 * simple_write_to_buffer - copy data from user space to the buffer
590 * @to: the buffer to write to
591 * @available: the size of the buffer
592 * @ppos: the current position in the buffer
593 * @from: the user space buffer to read from
594 * @count: the maximum number of bytes to read
595 *
596 * The simple_write_to_buffer() function reads up to @count bytes from the user
597 * space address starting at @from into the buffer @to at offset @ppos.
598 *
599 * On success, the number of bytes written is returned and the offset @ppos is
600 * advanced by this number, or negative value is returned on error.
601 **/
602ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
603 const void __user *from, size_t count)
604{
605 loff_t pos = *ppos;
606 size_t res;
607
608 if (pos < 0)
609 return -EINVAL;
610 if (pos >= available || !count)
611 return 0;
612 if (count > available - pos)
613 count = available - pos;
614 res = copy_from_user(to + pos, from, count);
615 if (res == count)
616 return -EFAULT;
617 count -= res;
618 *ppos = pos + count;
619 return count;
620}
621
622/**
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700623 * memory_read_from_buffer - copy data from the buffer
624 * @to: the kernel space buffer to read to
625 * @count: the maximum number of bytes to read
626 * @ppos: the current position in the buffer
627 * @from: the buffer to read from
628 * @available: the size of the buffer
629 *
630 * The memory_read_from_buffer() function reads up to @count bytes from the
631 * buffer @from at offset @ppos into the kernel space address starting at @to.
632 *
633 * On success, the number of bytes read is returned and the offset @ppos is
634 * advanced by this number, or negative value is returned on error.
635 **/
Akinobu Mita93b07112008-06-05 22:46:21 -0700636ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
637 const void *from, size_t available)
638{
639 loff_t pos = *ppos;
640
641 if (pos < 0)
642 return -EINVAL;
643 if (pos >= available)
644 return 0;
645 if (count > available - pos)
646 count = available - pos;
647 memcpy(to, from + pos, count);
648 *ppos = pos + count;
649
650 return count;
651}
652
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653/*
654 * Transaction based IO.
655 * The file expects a single write which triggers the transaction, and then
656 * possibly a read which collects the result - which is stored in a
657 * file-local buffer.
658 */
Ingo Molnar76791ab2009-03-25 16:48:35 +0100659
660void simple_transaction_set(struct file *file, size_t n)
661{
662 struct simple_transaction_argresp *ar = file->private_data;
663
664 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
665
666 /*
667 * The barrier ensures that ar->size will really remain zero until
668 * ar->data is ready for reading.
669 */
670 smp_mb();
671 ar->size = n;
672}
673
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
675{
676 struct simple_transaction_argresp *ar;
677 static DEFINE_SPINLOCK(simple_transaction_lock);
678
679 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
680 return ERR_PTR(-EFBIG);
681
682 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
683 if (!ar)
684 return ERR_PTR(-ENOMEM);
685
686 spin_lock(&simple_transaction_lock);
687
688 /* only one write allowed per open */
689 if (file->private_data) {
690 spin_unlock(&simple_transaction_lock);
691 free_page((unsigned long)ar);
692 return ERR_PTR(-EBUSY);
693 }
694
695 file->private_data = ar;
696
697 spin_unlock(&simple_transaction_lock);
698
699 if (copy_from_user(ar->data, buf, size))
700 return ERR_PTR(-EFAULT);
701
702 return ar->data;
703}
704
705ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
706{
707 struct simple_transaction_argresp *ar = file->private_data;
708
709 if (!ar)
710 return 0;
711 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
712}
713
714int simple_transaction_release(struct inode *inode, struct file *file)
715{
716 free_page((unsigned long)file->private_data);
717 return 0;
718}
719
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200720/* Simple attribute files */
721
722struct simple_attr {
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800723 int (*get)(void *, u64 *);
724 int (*set)(void *, u64);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200725 char get_buf[24]; /* enough to store a u64 and "\n\0" */
726 char set_buf[24];
727 void *data;
728 const char *fmt; /* format for read operation */
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800729 struct mutex mutex; /* protects access to these buffers */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200730};
731
732/* simple_attr_open is called by an actual attribute open file operation
733 * to set the attribute specific access operations. */
734int simple_attr_open(struct inode *inode, struct file *file,
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800735 int (*get)(void *, u64 *), int (*set)(void *, u64),
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200736 const char *fmt)
737{
738 struct simple_attr *attr;
739
740 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
741 if (!attr)
742 return -ENOMEM;
743
744 attr->get = get;
745 attr->set = set;
Theodore Ts'o8e18e292006-09-27 01:50:46 -0700746 attr->data = inode->i_private;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200747 attr->fmt = fmt;
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800748 mutex_init(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200749
750 file->private_data = attr;
751
752 return nonseekable_open(inode, file);
753}
754
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800755int simple_attr_release(struct inode *inode, struct file *file)
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200756{
757 kfree(file->private_data);
758 return 0;
759}
760
761/* read from the buffer that is filled with the get function */
762ssize_t simple_attr_read(struct file *file, char __user *buf,
763 size_t len, loff_t *ppos)
764{
765 struct simple_attr *attr;
766 size_t size;
767 ssize_t ret;
768
769 attr = file->private_data;
770
771 if (!attr->get)
772 return -EACCES;
773
Christoph Hellwig92613032008-02-08 04:20:27 -0800774 ret = mutex_lock_interruptible(&attr->mutex);
775 if (ret)
776 return ret;
777
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800778 if (*ppos) { /* continued read */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200779 size = strlen(attr->get_buf);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800780 } else { /* first read */
781 u64 val;
782 ret = attr->get(attr->data, &val);
783 if (ret)
784 goto out;
785
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200786 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800787 attr->fmt, (unsigned long long)val);
788 }
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200789
790 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800791out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800792 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200793 return ret;
794}
795
796/* interpret the buffer as a number to call the set function with */
797ssize_t simple_attr_write(struct file *file, const char __user *buf,
798 size_t len, loff_t *ppos)
799{
800 struct simple_attr *attr;
801 u64 val;
802 size_t size;
803 ssize_t ret;
804
805 attr = file->private_data;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200806 if (!attr->set)
807 return -EACCES;
808
Christoph Hellwig92613032008-02-08 04:20:27 -0800809 ret = mutex_lock_interruptible(&attr->mutex);
810 if (ret)
811 return ret;
812
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200813 ret = -EFAULT;
814 size = min(sizeof(attr->set_buf) - 1, len);
815 if (copy_from_user(attr->set_buf, buf, size))
816 goto out;
817
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200818 attr->set_buf[size] = '\0';
819 val = simple_strtol(attr->set_buf, NULL, 0);
Wu Fengguang05cc0ce2009-09-18 13:06:03 -0700820 ret = attr->set(attr->data, val);
821 if (ret == 0)
822 ret = len; /* on success, claim we got the whole input */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200823out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800824 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200825 return ret;
826}
827
Christoph Hellwig25961102007-10-21 16:42:05 -0700828/**
829 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
830 * @sb: filesystem to do the file handle conversion on
831 * @fid: file handle to convert
832 * @fh_len: length of the file handle in bytes
833 * @fh_type: type of file handle
834 * @get_inode: filesystem callback to retrieve inode
835 *
836 * This function decodes @fid as long as it has one of the well-known
837 * Linux filehandle types and calls @get_inode on it to retrieve the
838 * inode for the object specified in the file handle.
839 */
840struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
841 int fh_len, int fh_type, struct inode *(*get_inode)
842 (struct super_block *sb, u64 ino, u32 gen))
843{
844 struct inode *inode = NULL;
845
846 if (fh_len < 2)
847 return NULL;
848
849 switch (fh_type) {
850 case FILEID_INO32_GEN:
851 case FILEID_INO32_GEN_PARENT:
852 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
853 break;
854 }
855
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +0200856 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -0700857}
858EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
859
860/**
861 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
862 * @sb: filesystem to do the file handle conversion on
863 * @fid: file handle to convert
864 * @fh_len: length of the file handle in bytes
865 * @fh_type: type of file handle
866 * @get_inode: filesystem callback to retrieve inode
867 *
868 * This function decodes @fid as long as it has one of the well-known
869 * Linux filehandle types and calls @get_inode on it to retrieve the
870 * inode for the _parent_ object specified in the file handle if it
871 * is specified in the file handle, or NULL otherwise.
872 */
873struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
874 int fh_len, int fh_type, struct inode *(*get_inode)
875 (struct super_block *sb, u64 ino, u32 gen))
876{
877 struct inode *inode = NULL;
878
879 if (fh_len <= 2)
880 return NULL;
881
882 switch (fh_type) {
883 case FILEID_INO32_GEN_PARENT:
884 inode = get_inode(sb, fid->i32.parent_ino,
885 (fh_len > 3 ? fid->i32.parent_gen : 0));
886 break;
887 }
888
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +0200889 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -0700890}
891EXPORT_SYMBOL_GPL(generic_fh_to_parent);
892
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200893/**
894 * generic_file_fsync - generic fsync implementation for simple filesystems
895 * @file: file to synchronize
896 * @datasync: only synchronize essential metadata if true
897 *
898 * This is a generic implementation of the fsync method for simple
899 * filesystems which track all non-inode metadata in the buffers list
900 * hanging off the address_space structure.
901 */
902int generic_file_fsync(struct file *file, int datasync)
Al Virod5aacad2009-06-07 14:56:44 -0400903{
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200904 struct inode *inode = file->f_mapping->host;
Al Virod5aacad2009-06-07 14:56:44 -0400905 int err;
906 int ret;
907
908 ret = sync_mapping_buffers(inode->i_mapping);
909 if (!(inode->i_state & I_DIRTY))
910 return ret;
911 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
912 return ret;
913
Christoph Hellwigc37650162010-10-06 10:48:20 +0200914 err = sync_inode_metadata(inode, 1);
Al Virod5aacad2009-06-07 14:56:44 -0400915 if (ret == 0)
916 ret = err;
917 return ret;
918}
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200919EXPORT_SYMBOL(generic_file_fsync);
920
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -0700921/**
922 * generic_check_addressable - Check addressability of file system
923 * @blocksize_bits: log of file system block size
924 * @num_blocks: number of blocks in file system
925 *
926 * Determine whether a file system with @num_blocks blocks (and a
927 * block size of 2**@blocksize_bits) is addressable by the sector_t
928 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
929 */
930int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
931{
932 u64 last_fs_block = num_blocks - 1;
Joel Beckera33f13e2010-08-16 12:10:17 -0700933 u64 last_fs_page =
934 last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -0700935
936 if (unlikely(num_blocks == 0))
937 return 0;
938
939 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
940 return -EINVAL;
941
Joel Beckera33f13e2010-08-16 12:10:17 -0700942 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
943 (last_fs_page > (pgoff_t)(~0ULL))) {
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -0700944 return -EFBIG;
945 }
946 return 0;
947}
948EXPORT_SYMBOL(generic_check_addressable);
949
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200950/*
951 * No-op implementation of ->fsync for in-memory filesystems.
952 */
953int noop_fsync(struct file *file, int datasync)
954{
955 return 0;
956}
Al Virod5aacad2009-06-07 14:56:44 -0400957
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958EXPORT_SYMBOL(dcache_dir_close);
959EXPORT_SYMBOL(dcache_dir_lseek);
960EXPORT_SYMBOL(dcache_dir_open);
961EXPORT_SYMBOL(dcache_readdir);
962EXPORT_SYMBOL(generic_read_dir);
Al Viro51139ad2010-07-25 23:47:46 +0400963EXPORT_SYMBOL(mount_pseudo);
Nick Pigginafddba42007-10-16 01:25:01 -0700964EXPORT_SYMBOL(simple_write_begin);
965EXPORT_SYMBOL(simple_write_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966EXPORT_SYMBOL(simple_dir_inode_operations);
967EXPORT_SYMBOL(simple_dir_operations);
968EXPORT_SYMBOL(simple_empty);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969EXPORT_SYMBOL(simple_fill_super);
970EXPORT_SYMBOL(simple_getattr);
971EXPORT_SYMBOL(simple_link);
972EXPORT_SYMBOL(simple_lookup);
973EXPORT_SYMBOL(simple_pin_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974EXPORT_SYMBOL(simple_readpage);
975EXPORT_SYMBOL(simple_release_fs);
976EXPORT_SYMBOL(simple_rename);
977EXPORT_SYMBOL(simple_rmdir);
978EXPORT_SYMBOL(simple_statfs);
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200979EXPORT_SYMBOL(noop_fsync);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980EXPORT_SYMBOL(simple_unlink);
981EXPORT_SYMBOL(simple_read_from_buffer);
Jiri Slaby6a727b42010-05-01 23:51:22 +0200982EXPORT_SYMBOL(simple_write_to_buffer);
Akinobu Mita93b07112008-06-05 22:46:21 -0700983EXPORT_SYMBOL(memory_read_from_buffer);
Ingo Molnar76791ab2009-03-25 16:48:35 +0100984EXPORT_SYMBOL(simple_transaction_set);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985EXPORT_SYMBOL(simple_transaction_get);
986EXPORT_SYMBOL(simple_transaction_read);
987EXPORT_SYMBOL(simple_transaction_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200988EXPORT_SYMBOL_GPL(simple_attr_open);
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800989EXPORT_SYMBOL_GPL(simple_attr_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200990EXPORT_SYMBOL_GPL(simple_attr_read);
991EXPORT_SYMBOL_GPL(simple_attr_write);