blob: 5025563e7379675b7df2834427aabb0800f4b740 [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>
8#include <linux/mount.h>
9#include <linux/vfs.h>
10#include <asm/uaccess.h>
11
12int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
13 struct kstat *stat)
14{
15 struct inode *inode = dentry->d_inode;
16 generic_fillattr(inode, stat);
17 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
18 return 0;
19}
20
21int simple_statfs(struct super_block *sb, struct kstatfs *buf)
22{
23 buf->f_type = sb->s_magic;
24 buf->f_bsize = PAGE_CACHE_SIZE;
25 buf->f_namelen = NAME_MAX;
26 return 0;
27}
28
29/*
30 * Retaining negative dentries for an in-memory filesystem just wastes
31 * memory and lookup time: arrange for them to be deleted immediately.
32 */
33static int simple_delete_dentry(struct dentry *dentry)
34{
35 return 1;
36}
37
38/*
39 * Lookup the data. This is trivial - if the dentry didn't already
40 * exist, we know it is negative. Set d_op to delete negative dentries.
41 */
42struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
43{
44 static struct dentry_operations simple_dentry_operations = {
45 .d_delete = simple_delete_dentry,
46 };
47
48 if (dentry->d_name.len > NAME_MAX)
49 return ERR_PTR(-ENAMETOOLONG);
50 dentry->d_op = &simple_dentry_operations;
51 d_add(dentry, NULL);
52 return NULL;
53}
54
55int simple_sync_file(struct file * file, struct dentry *dentry, int datasync)
56{
57 return 0;
58}
59
60int dcache_dir_open(struct inode *inode, struct file *file)
61{
62 static struct qstr cursor_name = {.len = 1, .name = "."};
63
64 file->private_data = d_alloc(file->f_dentry, &cursor_name);
65
66 return file->private_data ? 0 : -ENOMEM;
67}
68
69int dcache_dir_close(struct inode *inode, struct file *file)
70{
71 dput(file->private_data);
72 return 0;
73}
74
75loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
76{
77 down(&file->f_dentry->d_inode->i_sem);
78 switch (origin) {
79 case 1:
80 offset += file->f_pos;
81 case 0:
82 if (offset >= 0)
83 break;
84 default:
85 up(&file->f_dentry->d_inode->i_sem);
86 return -EINVAL;
87 }
88 if (offset != file->f_pos) {
89 file->f_pos = offset;
90 if (file->f_pos >= 2) {
91 struct list_head *p;
92 struct dentry *cursor = file->private_data;
93 loff_t n = file->f_pos - 2;
94
95 spin_lock(&dcache_lock);
96 list_del(&cursor->d_child);
97 p = file->f_dentry->d_subdirs.next;
98 while (n && p != &file->f_dentry->d_subdirs) {
99 struct dentry *next;
100 next = list_entry(p, struct dentry, d_child);
101 if (!d_unhashed(next) && next->d_inode)
102 n--;
103 p = p->next;
104 }
105 list_add_tail(&cursor->d_child, p);
106 spin_unlock(&dcache_lock);
107 }
108 }
109 up(&file->f_dentry->d_inode->i_sem);
110 return offset;
111}
112
113/* Relationship between i_mode and the DT_xxx types */
114static inline unsigned char dt_type(struct inode *inode)
115{
116 return (inode->i_mode >> 12) & 15;
117}
118
119/*
120 * Directory is locked and all positive dentries in it are safe, since
121 * for ramfs-type trees they can't go away without unlink() or rmdir(),
122 * both impossible due to the lock on directory.
123 */
124
125int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
126{
127 struct dentry *dentry = filp->f_dentry;
128 struct dentry *cursor = filp->private_data;
129 struct list_head *p, *q = &cursor->d_child;
130 ino_t ino;
131 int i = filp->f_pos;
132
133 switch (i) {
134 case 0:
135 ino = dentry->d_inode->i_ino;
136 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
137 break;
138 filp->f_pos++;
139 i++;
140 /* fallthrough */
141 case 1:
142 ino = parent_ino(dentry);
143 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
144 break;
145 filp->f_pos++;
146 i++;
147 /* fallthrough */
148 default:
149 spin_lock(&dcache_lock);
150 if (filp->f_pos == 2) {
151 list_del(q);
152 list_add(q, &dentry->d_subdirs);
153 }
154 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
155 struct dentry *next;
156 next = list_entry(p, struct dentry, d_child);
157 if (d_unhashed(next) || !next->d_inode)
158 continue;
159
160 spin_unlock(&dcache_lock);
161 if (filldir(dirent, next->d_name.name, next->d_name.len, filp->f_pos, next->d_inode->i_ino, dt_type(next->d_inode)) < 0)
162 return 0;
163 spin_lock(&dcache_lock);
164 /* next is still alive */
165 list_del(q);
166 list_add(q, p);
167 p = q;
168 filp->f_pos++;
169 }
170 spin_unlock(&dcache_lock);
171 }
172 return 0;
173}
174
175ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
176{
177 return -EISDIR;
178}
179
180struct file_operations simple_dir_operations = {
181 .open = dcache_dir_open,
182 .release = dcache_dir_close,
183 .llseek = dcache_dir_lseek,
184 .read = generic_read_dir,
185 .readdir = dcache_readdir,
186};
187
188struct inode_operations simple_dir_inode_operations = {
189 .lookup = simple_lookup,
190};
191
192/*
193 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
194 * will never be mountable)
195 */
196struct super_block *
197get_sb_pseudo(struct file_system_type *fs_type, char *name,
198 struct super_operations *ops, unsigned long magic)
199{
200 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
201 static struct super_operations default_ops = {.statfs = simple_statfs};
202 struct dentry *dentry;
203 struct inode *root;
204 struct qstr d_name = {.name = name, .len = strlen(name)};
205
206 if (IS_ERR(s))
207 return s;
208
209 s->s_flags = MS_NOUSER;
210 s->s_maxbytes = ~0ULL;
211 s->s_blocksize = 1024;
212 s->s_blocksize_bits = 10;
213 s->s_magic = magic;
214 s->s_op = ops ? ops : &default_ops;
215 s->s_time_gran = 1;
216 root = new_inode(s);
217 if (!root)
218 goto Enomem;
219 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
220 root->i_uid = root->i_gid = 0;
221 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
222 dentry = d_alloc(NULL, &d_name);
223 if (!dentry) {
224 iput(root);
225 goto Enomem;
226 }
227 dentry->d_sb = s;
228 dentry->d_parent = dentry;
229 d_instantiate(dentry, root);
230 s->s_root = dentry;
231 s->s_flags |= MS_ACTIVE;
232 return s;
233
234Enomem:
235 up_write(&s->s_umount);
236 deactivate_super(s);
237 return ERR_PTR(-ENOMEM);
238}
239
240int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
241{
242 struct inode *inode = old_dentry->d_inode;
243
244 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
245 inode->i_nlink++;
246 atomic_inc(&inode->i_count);
247 dget(dentry);
248 d_instantiate(dentry, inode);
249 return 0;
250}
251
252static inline int simple_positive(struct dentry *dentry)
253{
254 return dentry->d_inode && !d_unhashed(dentry);
255}
256
257int simple_empty(struct dentry *dentry)
258{
259 struct dentry *child;
260 int ret = 0;
261
262 spin_lock(&dcache_lock);
263 list_for_each_entry(child, &dentry->d_subdirs, d_child)
264 if (simple_positive(child))
265 goto out;
266 ret = 1;
267out:
268 spin_unlock(&dcache_lock);
269 return ret;
270}
271
272int simple_unlink(struct inode *dir, struct dentry *dentry)
273{
274 struct inode *inode = dentry->d_inode;
275
276 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
277 inode->i_nlink--;
278 dput(dentry);
279 return 0;
280}
281
282int simple_rmdir(struct inode *dir, struct dentry *dentry)
283{
284 if (!simple_empty(dentry))
285 return -ENOTEMPTY;
286
287 dentry->d_inode->i_nlink--;
288 simple_unlink(dir, dentry);
289 dir->i_nlink--;
290 return 0;
291}
292
293int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
294 struct inode *new_dir, struct dentry *new_dentry)
295{
296 struct inode *inode = old_dentry->d_inode;
297 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
298
299 if (!simple_empty(new_dentry))
300 return -ENOTEMPTY;
301
302 if (new_dentry->d_inode) {
303 simple_unlink(new_dir, new_dentry);
304 if (they_are_dirs)
305 old_dir->i_nlink--;
306 } else if (they_are_dirs) {
307 old_dir->i_nlink--;
308 new_dir->i_nlink++;
309 }
310
311 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
312 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
313
314 return 0;
315}
316
317int simple_readpage(struct file *file, struct page *page)
318{
319 void *kaddr;
320
321 if (PageUptodate(page))
322 goto out;
323
324 kaddr = kmap_atomic(page, KM_USER0);
325 memset(kaddr, 0, PAGE_CACHE_SIZE);
326 kunmap_atomic(kaddr, KM_USER0);
327 flush_dcache_page(page);
328 SetPageUptodate(page);
329out:
330 unlock_page(page);
331 return 0;
332}
333
334int simple_prepare_write(struct file *file, struct page *page,
335 unsigned from, unsigned to)
336{
337 if (!PageUptodate(page)) {
338 if (to - from != PAGE_CACHE_SIZE) {
339 void *kaddr = kmap_atomic(page, KM_USER0);
340 memset(kaddr, 0, from);
341 memset(kaddr + to, 0, PAGE_CACHE_SIZE - to);
342 flush_dcache_page(page);
343 kunmap_atomic(kaddr, KM_USER0);
344 }
345 SetPageUptodate(page);
346 }
347 return 0;
348}
349
350int simple_commit_write(struct file *file, struct page *page,
351 unsigned offset, unsigned to)
352{
353 struct inode *inode = page->mapping->host;
354 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
355
356 /*
357 * No need to use i_size_read() here, the i_size
358 * cannot change under us because we hold the i_sem.
359 */
360 if (pos > inode->i_size)
361 i_size_write(inode, pos);
362 set_page_dirty(page);
363 return 0;
364}
365
366int simple_fill_super(struct super_block *s, int magic, struct tree_descr *files)
367{
368 static struct super_operations s_ops = {.statfs = simple_statfs};
369 struct inode *inode;
370 struct dentry *root;
371 struct dentry *dentry;
372 int i;
373
374 s->s_blocksize = PAGE_CACHE_SIZE;
375 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
376 s->s_magic = magic;
377 s->s_op = &s_ops;
378 s->s_time_gran = 1;
379
380 inode = new_inode(s);
381 if (!inode)
382 return -ENOMEM;
383 inode->i_mode = S_IFDIR | 0755;
384 inode->i_uid = inode->i_gid = 0;
385 inode->i_blksize = PAGE_CACHE_SIZE;
386 inode->i_blocks = 0;
387 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
388 inode->i_op = &simple_dir_inode_operations;
389 inode->i_fop = &simple_dir_operations;
390 root = d_alloc_root(inode);
391 if (!root) {
392 iput(inode);
393 return -ENOMEM;
394 }
395 for (i = 0; !files->name || files->name[0]; i++, files++) {
396 if (!files->name)
397 continue;
398 dentry = d_alloc_name(root, files->name);
399 if (!dentry)
400 goto out;
401 inode = new_inode(s);
402 if (!inode)
403 goto out;
404 inode->i_mode = S_IFREG | files->mode;
405 inode->i_uid = inode->i_gid = 0;
406 inode->i_blksize = PAGE_CACHE_SIZE;
407 inode->i_blocks = 0;
408 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
409 inode->i_fop = files->ops;
410 inode->i_ino = i;
411 d_add(dentry, inode);
412 }
413 s->s_root = root;
414 return 0;
415out:
416 d_genocide(root);
417 dput(root);
418 return -ENOMEM;
419}
420
421static DEFINE_SPINLOCK(pin_fs_lock);
422
423int simple_pin_fs(char *name, struct vfsmount **mount, int *count)
424{
425 struct vfsmount *mnt = NULL;
426 spin_lock(&pin_fs_lock);
427 if (unlikely(!*mount)) {
428 spin_unlock(&pin_fs_lock);
429 mnt = do_kern_mount(name, 0, name, NULL);
430 if (IS_ERR(mnt))
431 return PTR_ERR(mnt);
432 spin_lock(&pin_fs_lock);
433 if (!*mount)
434 *mount = mnt;
435 }
436 mntget(*mount);
437 ++*count;
438 spin_unlock(&pin_fs_lock);
439 mntput(mnt);
440 return 0;
441}
442
443void simple_release_fs(struct vfsmount **mount, int *count)
444{
445 struct vfsmount *mnt;
446 spin_lock(&pin_fs_lock);
447 mnt = *mount;
448 if (!--*count)
449 *mount = NULL;
450 spin_unlock(&pin_fs_lock);
451 mntput(mnt);
452}
453
454ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
455 const void *from, size_t available)
456{
457 loff_t pos = *ppos;
458 if (pos < 0)
459 return -EINVAL;
460 if (pos >= available)
461 return 0;
462 if (count > available - pos)
463 count = available - pos;
464 if (copy_to_user(to, from + pos, count))
465 return -EFAULT;
466 *ppos = pos + count;
467 return count;
468}
469
470/*
471 * Transaction based IO.
472 * The file expects a single write which triggers the transaction, and then
473 * possibly a read which collects the result - which is stored in a
474 * file-local buffer.
475 */
476char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
477{
478 struct simple_transaction_argresp *ar;
479 static DEFINE_SPINLOCK(simple_transaction_lock);
480
481 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
482 return ERR_PTR(-EFBIG);
483
484 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
485 if (!ar)
486 return ERR_PTR(-ENOMEM);
487
488 spin_lock(&simple_transaction_lock);
489
490 /* only one write allowed per open */
491 if (file->private_data) {
492 spin_unlock(&simple_transaction_lock);
493 free_page((unsigned long)ar);
494 return ERR_PTR(-EBUSY);
495 }
496
497 file->private_data = ar;
498
499 spin_unlock(&simple_transaction_lock);
500
501 if (copy_from_user(ar->data, buf, size))
502 return ERR_PTR(-EFAULT);
503
504 return ar->data;
505}
506
507ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
508{
509 struct simple_transaction_argresp *ar = file->private_data;
510
511 if (!ar)
512 return 0;
513 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
514}
515
516int simple_transaction_release(struct inode *inode, struct file *file)
517{
518 free_page((unsigned long)file->private_data);
519 return 0;
520}
521
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200522/* Simple attribute files */
523
524struct simple_attr {
525 u64 (*get)(void *);
526 void (*set)(void *, u64);
527 char get_buf[24]; /* enough to store a u64 and "\n\0" */
528 char set_buf[24];
529 void *data;
530 const char *fmt; /* format for read operation */
531 struct semaphore sem; /* protects access to these buffers */
532};
533
534/* simple_attr_open is called by an actual attribute open file operation
535 * to set the attribute specific access operations. */
536int simple_attr_open(struct inode *inode, struct file *file,
537 u64 (*get)(void *), void (*set)(void *, u64),
538 const char *fmt)
539{
540 struct simple_attr *attr;
541
542 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
543 if (!attr)
544 return -ENOMEM;
545
546 attr->get = get;
547 attr->set = set;
548 attr->data = inode->u.generic_ip;
549 attr->fmt = fmt;
550 init_MUTEX(&attr->sem);
551
552 file->private_data = attr;
553
554 return nonseekable_open(inode, file);
555}
556
557int simple_attr_close(struct inode *inode, struct file *file)
558{
559 kfree(file->private_data);
560 return 0;
561}
562
563/* read from the buffer that is filled with the get function */
564ssize_t simple_attr_read(struct file *file, char __user *buf,
565 size_t len, loff_t *ppos)
566{
567 struct simple_attr *attr;
568 size_t size;
569 ssize_t ret;
570
571 attr = file->private_data;
572
573 if (!attr->get)
574 return -EACCES;
575
576 down(&attr->sem);
577 if (*ppos) /* continued read */
578 size = strlen(attr->get_buf);
579 else /* first read */
580 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
581 attr->fmt,
582 (unsigned long long)attr->get(attr->data));
583
584 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
585 up(&attr->sem);
586 return ret;
587}
588
589/* interpret the buffer as a number to call the set function with */
590ssize_t simple_attr_write(struct file *file, const char __user *buf,
591 size_t len, loff_t *ppos)
592{
593 struct simple_attr *attr;
594 u64 val;
595 size_t size;
596 ssize_t ret;
597
598 attr = file->private_data;
599
600 if (!attr->set)
601 return -EACCES;
602
603 down(&attr->sem);
604 ret = -EFAULT;
605 size = min(sizeof(attr->set_buf) - 1, len);
606 if (copy_from_user(attr->set_buf, buf, size))
607 goto out;
608
609 ret = len; /* claim we got the whole input */
610 attr->set_buf[size] = '\0';
611 val = simple_strtol(attr->set_buf, NULL, 0);
612 attr->set(attr->data, val);
613out:
614 up(&attr->sem);
615 return ret;
616}
617
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618EXPORT_SYMBOL(dcache_dir_close);
619EXPORT_SYMBOL(dcache_dir_lseek);
620EXPORT_SYMBOL(dcache_dir_open);
621EXPORT_SYMBOL(dcache_readdir);
622EXPORT_SYMBOL(generic_read_dir);
623EXPORT_SYMBOL(get_sb_pseudo);
624EXPORT_SYMBOL(simple_commit_write);
625EXPORT_SYMBOL(simple_dir_inode_operations);
626EXPORT_SYMBOL(simple_dir_operations);
627EXPORT_SYMBOL(simple_empty);
628EXPORT_SYMBOL(d_alloc_name);
629EXPORT_SYMBOL(simple_fill_super);
630EXPORT_SYMBOL(simple_getattr);
631EXPORT_SYMBOL(simple_link);
632EXPORT_SYMBOL(simple_lookup);
633EXPORT_SYMBOL(simple_pin_fs);
634EXPORT_SYMBOL(simple_prepare_write);
635EXPORT_SYMBOL(simple_readpage);
636EXPORT_SYMBOL(simple_release_fs);
637EXPORT_SYMBOL(simple_rename);
638EXPORT_SYMBOL(simple_rmdir);
639EXPORT_SYMBOL(simple_statfs);
640EXPORT_SYMBOL(simple_sync_file);
641EXPORT_SYMBOL(simple_unlink);
642EXPORT_SYMBOL(simple_read_from_buffer);
643EXPORT_SYMBOL(simple_transaction_get);
644EXPORT_SYMBOL(simple_transaction_read);
645EXPORT_SYMBOL(simple_transaction_release);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200646EXPORT_SYMBOL_GPL(simple_attr_open);
647EXPORT_SYMBOL_GPL(simple_attr_close);
648EXPORT_SYMBOL_GPL(simple_attr_read);
649EXPORT_SYMBOL_GPL(simple_attr_write);