libfs.c en linux

  1 /*
  2  *      fs/libfs.c
  3  *      Library for filesystems writers.
  4  */
  5 
  6 #include <linux/blkdev.h>
  7 #include <linux/export.h>
  8 #include <linux/pagemap.h>
  9 #include <linux/slab.h>
 10 #include <linux/mount.h>
 11 #include <linux/vfs.h>
 12 #include <linux/quotaops.h>
 13 #include <linux/mutex.h>
 14 #include <linux/namei.h>
 15 #include <linux/exportfs.h>
 16 #include <linux/writeback.h>
 17 #include <linux/buffer_head.h> /* sync_mapping_buffers */
 18 
 19 #include <asm/uaccess.h>
 20 
 21 #include "internal.h"
 22 
 23 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
 24                    struct kstat *stat)
 25 {
 26         struct inode *inode = d_inode(dentry);
 27         generic_fillattr(inode, stat);
 28         stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
 29         return 0;
 30 }
 31 EXPORT_SYMBOL(simple_getattr);
 32 
 33 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
 34 {
 35         buf->f_type = dentry->d_sb->s_magic;
 36         buf->f_bsize = PAGE_SIZE;
 37         buf->f_namelen = NAME_MAX;
 38         return 0;
 39 }
 40 EXPORT_SYMBOL(simple_statfs);
 41 
 42 /*
 43  * Retaining negative dentries for an in-memory filesystem just wastes
 44  * memory and lookup time: arrange for them to be deleted immediately.
 45  */
 46 int always_delete_dentry(const struct dentry *dentry)
 47 {
 48         return 1;
 49 }
 50 EXPORT_SYMBOL(always_delete_dentry);
 51 
 52 const struct dentry_operations simple_dentry_operations = {
 53         .d_delete = always_delete_dentry,
 54 };
 55 EXPORT_SYMBOL(simple_dentry_operations);
 56 
 57 /*
 58  * Lookup the data. This is trivial - if the dentry didn't already
 59  * exist, we know it is negative.  Set d_op to delete negative dentries.
 60  */
 61 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
 62 {
 63         if (dentry->d_name.len > NAME_MAX)
 64                 return ERR_PTR(-ENAMETOOLONG);
 65         if (!dentry->d_sb->s_d_op)
 66                 d_set_d_op(dentry, &simple_dentry_operations);
 67         d_add(dentry, NULL);
 68         return NULL;
 69 }
 70 EXPORT_SYMBOL(simple_lookup);
 71 
 72 int dcache_dir_open(struct inode *inode, struct file *file)
 73 {
 74         static struct qstr cursor_name = QSTR_INIT(".", 1);
 75 
 76         file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
 77 
 78         return file->private_data ? 0 : -ENOMEM;
 79 }
 80 EXPORT_SYMBOL(dcache_dir_open);
 81 
 82 int dcache_dir_close(struct inode *inode, struct file *file)
 83 {
 84         dput(file->private_data);
 85         return 0;
 86 }
 87 EXPORT_SYMBOL(dcache_dir_close);
 88 
 89 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
 90 {
 91         struct dentry *dentry = file->f_path.dentry;
 92         inode_lock(d_inode(dentry));
 93         switch (whence) {
 94                 case 1:
 95                         offset += file->f_pos;
 96                 case 0:
 97                         if (offset >= 0)
 98                                 break;
 99                 default:
100                         inode_unlock(d_inode(dentry));
101                         return -EINVAL;
102         }
103         if (offset != file->f_pos) {
104                 file->f_pos = offset;
105                 if (file->f_pos >= 2) {
106                         struct list_head *p;
107                         struct dentry *cursor = file->private_data;
108                         loff_t n = file->f_pos - 2;
109 
110                         spin_lock(&dentry->d_lock);
111                         /* d_lock not required for cursor */
112                         list_del(&cursor->d_child);
113                         p = dentry->d_subdirs.next;
114                         while (n && p != &dentry->d_subdirs) {
115                                 struct dentry *next;
116                                 next = list_entry(p, struct dentry, d_child);
117                                 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
118                                 if (simple_positive(next))
119                                         n--;
120                                 spin_unlock(&next->d_lock);
121                                 p = p->next;
122                         }
123                         list_add_tail(&cursor->d_child, p);
124                         spin_unlock(&dentry->d_lock);
125                 }
126         }
127         inode_unlock(d_inode(dentry));
128         return offset;
129 }
130 EXPORT_SYMBOL(dcache_dir_lseek);
131 
132 /* Relationship between i_mode and the DT_xxx types */
133 static inline unsigned char dt_type(struct inode *inode)
134 {
135         return (inode->i_mode >> 12) & 15;
136 }
137 
138 /*
139  * Directory is locked and all positive dentries in it are safe, since
140  * for ramfs-type trees they can't go away without unlink() or rmdir(),
141  * both impossible due to the lock on directory.
142  */
143 
144 int dcache_readdir(struct file *file, struct dir_context *ctx)
145 {
146         struct dentry *dentry = file->f_path.dentry;
147         struct dentry *cursor = file->private_data;
148         struct list_head *p, *q = &cursor->d_child;
149 
150         if (!dir_emit_dots(file, ctx))
151                 return 0;
152         spin_lock(&dentry->d_lock);
153         if (ctx->pos == 2)
154                 list_move(q, &dentry->d_subdirs);
155 
156         for (p = q->next; p != &dentry->d_subdirs; p = p->next) {
157                 struct dentry *next = list_entry(p, struct dentry, d_child);
158                 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
159                 if (!simple_positive(next)) {
160                         spin_unlock(&next->d_lock);
161                         continue;
162                 }
163 
164                 spin_unlock(&next->d_lock);
165                 spin_unlock(&dentry->d_lock);
166                 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
167                               d_inode(next)->i_ino, dt_type(d_inode(next))))
168                         return 0;
169                 spin_lock(&dentry->d_lock);
170                 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
171                 /* next is still alive */
172                 list_move(q, p);
173                 spin_unlock(&next->d_lock);
174                 p = q;
175                 ctx->pos++;
176         }
177         spin_unlock(&dentry->d_lock);
178         return 0;
179 }
180 EXPORT_SYMBOL(dcache_readdir);
181 
182 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
183 {
184         return -EISDIR;
185 }
186 EXPORT_SYMBOL(generic_read_dir);
187 
188 const struct file_operations simple_dir_operations = {
189         .open           = dcache_dir_open,
190         .release        = dcache_dir_close,
191         .llseek         = dcache_dir_lseek,
192         .read           = generic_read_dir,
193         .iterate        = dcache_readdir,
194         .fsync          = noop_fsync,
195 };
196 EXPORT_SYMBOL(simple_dir_operations);
197 
198 const struct inode_operations simple_dir_inode_operations = {
199         .lookup         = simple_lookup,
200 };
201 EXPORT_SYMBOL(simple_dir_inode_operations);
202 
203 static const struct super_operations simple_super_operations = {
204         .statfs         = simple_statfs,
205 };
206 
207 /*
208  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
209  * will never be mountable)
210  */
211 struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
212         const struct super_operations *ops,
213         const struct dentry_operations *dops, unsigned long magic)
214 {
215         struct super_block *s;
216         struct dentry *dentry;
217         struct inode *root;
218         struct qstr d_name = QSTR_INIT(name, strlen(name));
219 
220         s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
221         if (IS_ERR(s))
222                 return ERR_CAST(s);
223 
224         s->s_maxbytes = MAX_LFS_FILESIZE;
225         s->s_blocksize = PAGE_SIZE;
226         s->s_blocksize_bits = PAGE_SHIFT;
227         s->s_magic = magic;
228         s->s_op = ops ? ops : &simple_super_operations;
229         s->s_time_gran = 1;
230         root = new_inode(s);
231         if (!root)
232                 goto Enomem;
233         /*
234          * since this is the first inode, make it number 1. New inodes created
235          * after this must take care not to collide with it (by passing
236          * max_reserved of 1 to iunique).
237          */
238         root->i_ino = 1;
239         root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
240         root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
241         dentry = __d_alloc(s, &d_name);
242         if (!dentry) {
243                 iput(root);
244                 goto Enomem;
245         }
246         d_instantiate(dentry, root);
247         s->s_root = dentry;
248         s->s_d_op = dops;
249         s->s_flags |= MS_ACTIVE;
250         return dget(s->s_root);
251 
252 Enomem:
253         deactivate_locked_super(s);
254         return ERR_PTR(-ENOMEM);
255 }
256 EXPORT_SYMBOL(mount_pseudo);
257 
258 int simple_open(struct inode *inode, struct file *file)
259 {
260         if (inode->i_private)
261                 file->private_data = inode->i_private;
262         return 0;
263 }
264 EXPORT_SYMBOL(simple_open);
265 
266 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
267 {
268         struct inode *inode = d_inode(old_dentry);
269 
270         inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
271         inc_nlink(inode);
272         ihold(inode);
273         dget(dentry);
274         d_instantiate(dentry, inode);
275         return 0;
276 }
277 EXPORT_SYMBOL(simple_link);
278 
279 int simple_empty(struct dentry *dentry)
280 {
281         struct dentry *child;
282         int ret = 0;
283 
284         spin_lock(&dentry->d_lock);
285         list_for_each_entry(child, &dentry->d_subdirs, d_child) {
286                 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
287                 if (simple_positive(child)) {
288                         spin_unlock(&child->d_lock);
289                         goto out;
290                 }
291                 spin_unlock(&child->d_lock);
292         }
293         ret = 1;
294 out:
295         spin_unlock(&dentry->d_lock);
296         return ret;
297 }
298 EXPORT_SYMBOL(simple_empty);
299 
300 int simple_unlink(struct inode *dir, struct dentry *dentry)
301 {
302         struct inode *inode = d_inode(dentry);
303 
304         inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
305         drop_nlink(inode);
306         dput(dentry);
307         return 0;
308 }
309 EXPORT_SYMBOL(simple_unlink);
310 
311 int simple_rmdir(struct inode *dir, struct dentry *dentry)
312 {
313         if (!simple_empty(dentry))
314                 return -ENOTEMPTY;
315 
316         drop_nlink(d_inode(dentry));
317         simple_unlink(dir, dentry);
318         drop_nlink(dir);
319         return 0;
320 }
321 EXPORT_SYMBOL(simple_rmdir);
322 
323 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
324                 struct inode *new_dir, struct dentry *new_dentry)
325 {
326         struct inode *inode = d_inode(old_dentry);
327         int they_are_dirs = d_is_dir(old_dentry);
328 
329         if (!simple_empty(new_dentry))
330                 return -ENOTEMPTY;
331 
332         if (d_really_is_positive(new_dentry)) {
333                 simple_unlink(new_dir, new_dentry);
334                 if (they_are_dirs) {
335                         drop_nlink(d_inode(new_dentry));
336                         drop_nlink(old_dir);
337                 }
338         } else if (they_are_dirs) {
339                 drop_nlink(old_dir);
340                 inc_nlink(new_dir);
341         }
342 
343         old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
344                 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
345 
346         return 0;
347 }
348 EXPORT_SYMBOL(simple_rename);
349 
350 /**
351  * simple_setattr - setattr for simple filesystem
352  * @dentry: dentry
353  * @iattr: iattr structure
354  *
355  * Returns 0 on success, -error on failure.
356  *
357  * simple_setattr is a simple ->setattr implementation without a proper
358  * implementation of size changes.
359  *
360  * It can either be used for in-memory filesystems or special files
361  * on simple regular filesystems.  Anything that needs to change on-disk
362  * or wire state on size changes needs its own setattr method.
363  */
364 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
365 {
366         struct inode *inode = d_inode(dentry);
367         int error;
368 
369         error = inode_change_ok(inode, iattr);
370         if (error)
371                 return error;
372 
373         if (iattr->ia_valid & ATTR_SIZE)
374                 truncate_setsize(inode, iattr->ia_size);
375         setattr_copy(inode, iattr);
376         mark_inode_dirty(inode);
377         return 0;
378 }
379 EXPORT_SYMBOL(simple_setattr);
380 
381 int simple_readpage(struct file *file, struct page *page)
382 {
383         clear_highpage(page);
384         flush_dcache_page(page);
385         SetPageUptodate(page);
386         unlock_page(page);
387         return 0;
388 }
389 EXPORT_SYMBOL(simple_readpage);
390 
391 int simple_write_begin(struct file *file, struct address_space *mapping,
392                         loff_t pos, unsigned len, unsigned flags,
393                         struct page **pagep, void **fsdata)
394 {
395         struct page *page;
396         pgoff_t index;
397 
398         index = pos >> PAGE_SHIFT;
399 
400         page = grab_cache_page_write_begin(mapping, index, flags);
401         if (!page)
402                 return -ENOMEM;
403 
404         *pagep = page;
405 
406         if (!PageUptodate(page) && (len != PAGE_SIZE)) {
407                 unsigned from = pos & (PAGE_SIZE - 1);
408 
409                 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
410         }
411         return 0;
412 }
413 EXPORT_SYMBOL(simple_write_begin);
414 
415 /**
416  * simple_write_end - .write_end helper for non-block-device FSes
417  * @available: See .write_end of address_space_operations
418  * @file:               "
419  * @mapping:            "
420  * @pos:                "
421  * @len:                "
422  * @copied:             "
423  * @page:               "
424  * @fsdata:             "
425  *
426  * simple_write_end does the minimum needed for updating a page after writing is
427  * done. It has the same API signature as the .write_end of
428  * address_space_operations vector. So it can just be set onto .write_end for
429  * FSes that don't need any other processing. i_mutex is assumed to be held.
430  * Block based filesystems should use generic_write_end().
431  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
432  * is not called, so a filesystem that actually does store data in .write_inode
433  * should extend on what's done here with a call to mark_inode_dirty() in the
434  * case that i_size has changed.
435  */
436 int simple_write_end(struct file *file, struct address_space *mapping,
437                         loff_t pos, unsigned len, unsigned copied,
438                         struct page *page, void *fsdata)
439 {
440         struct inode *inode = page->mapping->host;
441         loff_t last_pos = pos + copied;
442 
443         /* zero the stale part of the page if we did a short copy */
444         if (copied < len) {
445                 unsigned from = pos & (PAGE_SIZE - 1);
446 
447                 zero_user(page, from + copied, len - copied);
448         }
449 
450         if (!PageUptodate(page))
451                 SetPageUptodate(page);
452         /*
453          * No need to use i_size_read() here, the i_size
454          * cannot change under us because we hold the i_mutex.
455          */
456         if (last_pos > inode->i_size)
457                 i_size_write(inode, last_pos);
458 
459         set_page_dirty(page);
460         unlock_page(page);
461         put_page(page);
462 
463         return copied;
464 }
465 EXPORT_SYMBOL(simple_write_end);
466 
467 /*
468  * the inodes created here are not hashed. If you use iunique to generate
469  * unique inode values later for this filesystem, then you must take care
470  * to pass it an appropriate max_reserved value to avoid collisions.
471  */
472 int simple_fill_super(struct super_block *s, unsigned long magic,
473                       struct tree_descr *files)
474 {
475         struct inode *inode;
476         struct dentry *root;
477         struct dentry *dentry;
478         int i;
479 
480         s->s_blocksize = PAGE_SIZE;
481         s->s_blocksize_bits = PAGE_SHIFT;
482         s->s_magic = magic;
483         s->s_op = &simple_super_operations;
484         s->s_time_gran = 1;
485 
486         inode = new_inode(s);
487         if (!inode)
488                 return -ENOMEM;
489         /*
490          * because the root inode is 1, the files array must not contain an
491          * entry at index 1
492          */
493         inode->i_ino = 1;
494         inode->i_mode = S_IFDIR | 0755;
495         inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
496         inode->i_op = &simple_dir_inode_operations;
497         inode->i_fop = &simple_dir_operations;
498         set_nlink(inode, 2);
499         root = d_make_root(inode);
500         if (!root)
501                 return -ENOMEM;
502         for (i = 0; !files->name || files->name[0]; i++, files++) {
503                 if (!files->name)
504                         continue;
505 
506                 /* warn if it tries to conflict with the root inode */
507                 if (unlikely(i == 1))
508                         printk(KERN_WARNING "%s: %s passed in a files array"
509                                 "with an index of 1!\n", __func__,
510                                 s->s_type->name);
511 
512                 dentry = d_alloc_name(root, files->name);
513                 if (!dentry)
514                         goto out;
515                 inode = new_inode(s);
516                 if (!inode) {
517                         dput(dentry);
518                         goto out;
519                 }
520                 inode->i_mode = S_IFREG | files->mode;
521                 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
522                 inode->i_fop = files->ops;
523                 inode->i_ino = i;
524                 d_add(dentry, inode);
525         }
526         s->s_root = root;
527         return 0;
528 out:
529         d_genocide(root);
530         shrink_dcache_parent(root);
531         dput(root);
532         return -ENOMEM;
533 }
534 EXPORT_SYMBOL(simple_fill_super);
535 
536 static DEFINE_SPINLOCK(pin_fs_lock);
537 
538 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
539 {
540         struct vfsmount *mnt = NULL;
541         spin_lock(&pin_fs_lock);
542         if (unlikely(!*mount)) {
543                 spin_unlock(&pin_fs_lock);
544                 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
545                 if (IS_ERR(mnt))
546                         return PTR_ERR(mnt);
547                 spin_lock(&pin_fs_lock);
548                 if (!*mount)
549                         *mount = mnt;
550         }
551         mntget(*mount);
552         ++*count;
553         spin_unlock(&pin_fs_lock);
554         mntput(mnt);
555         return 0;
556 }
557 EXPORT_SYMBOL(simple_pin_fs);
558 
559 void simple_release_fs(struct vfsmount **mount, int *count)
560 {
561         struct vfsmount *mnt;
562         spin_lock(&pin_fs_lock);
563         mnt = *mount;
564         if (!--*count)
565                 *mount = NULL;
566         spin_unlock(&pin_fs_lock);
567         mntput(mnt);
568 }
569 EXPORT_SYMBOL(simple_release_fs);
570 
571 /**
572  * simple_read_from_buffer - copy data from the buffer to user space
573  * @to: the user space buffer to read to
574  * @count: the maximum number of bytes to read
575  * @ppos: the current position in the buffer
576  * @from: the buffer to read from
577  * @available: the size of the buffer
578  *
579  * The simple_read_from_buffer() function reads up to @count bytes from the
580  * buffer @from at offset @ppos into the user space address starting at @to.
581  *
582  * On success, the number of bytes read is returned and the offset @ppos is
583  * advanced by this number, or negative value is returned on error.
584  **/
585 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
586                                 const void *from, size_t available)
587 {
588         loff_t pos = *ppos;
589         size_t ret;
590 
591         if (pos < 0)
592                 return -EINVAL;
593         if (pos >= available || !count)
594                 return 0;
595         if (count > available - pos)
596                 count = available - pos;
597         ret = copy_to_user(to, from + pos, count);
598         if (ret == count)
599                 return -EFAULT;
600         count -= ret;
601         *ppos = pos + count;
602         return count;
603 }
604 EXPORT_SYMBOL(simple_read_from_buffer);
605 
606 /**
607  * simple_write_to_buffer - copy data from user space to the buffer
608  * @to: the buffer to write to
609  * @available: the size of the buffer
610  * @ppos: the current position in the buffer
611  * @from: the user space buffer to read from
612  * @count: the maximum number of bytes to read
613  *
614  * The simple_write_to_buffer() function reads up to @count bytes from the user
615  * space address starting at @from into the buffer @to at offset @ppos.
616  *
617  * On success, the number of bytes written is returned and the offset @ppos is
618  * advanced by this number, or negative value is returned on error.
619  **/
620 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
621                 const void __user *from, size_t count)
622 {
623         loff_t pos = *ppos;
624         size_t res;
625 
626         if (pos < 0)
627                 return -EINVAL;
628         if (pos >= available || !count)
629                 return 0;
630         if (count > available - pos)
631                 count = available - pos;
632         res = copy_from_user(to + pos, from, count);
633         if (res == count)
634                 return -EFAULT;
635         count -= res;
636         *ppos = pos + count;
637         return count;
638 }
639 EXPORT_SYMBOL(simple_write_to_buffer);
640 
641 /**
642  * memory_read_from_buffer - copy data from the buffer
643  * @to: the kernel space buffer to read to
644  * @count: the maximum number of bytes to read
645  * @ppos: the current position in the buffer
646  * @from: the buffer to read from
647  * @available: the size of the buffer
648  *
649  * The memory_read_from_buffer() function reads up to @count bytes from the
650  * buffer @from at offset @ppos into the kernel space address starting at @to.
651  *
652  * On success, the number of bytes read is returned and the offset @ppos is
653  * advanced by this number, or negative value is returned on error.
654  **/
655 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
656                                 const void *from, size_t available)
657 {
658         loff_t pos = *ppos;
659 
660         if (pos < 0)
661                 return -EINVAL;
662         if (pos >= available)
663                 return 0;
664         if (count > available - pos)
665                 count = available - pos;
666         memcpy(to, from + pos, count);
667         *ppos = pos + count;
668 
669         return count;
670 }
671 EXPORT_SYMBOL(memory_read_from_buffer);
672 
673 /*
674  * Transaction based IO.
675  * The file expects a single write which triggers the transaction, and then
676  * possibly a read which collects the result - which is stored in a
677  * file-local buffer.
678  */
679 
680 void simple_transaction_set(struct file *file, size_t n)
681 {
682         struct simple_transaction_argresp *ar = file->private_data;
683 
684         BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
685 
686         /*
687          * The barrier ensures that ar->size will really remain zero until
688          * ar->data is ready for reading.
689          */
690         smp_mb();
691         ar->size = n;
692 }
693 EXPORT_SYMBOL(simple_transaction_set);
694 
695 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
696 {
697         struct simple_transaction_argresp *ar;
698         static DEFINE_SPINLOCK(simple_transaction_lock);
699 
700         if (size > SIMPLE_TRANSACTION_LIMIT - 1)
701                 return ERR_PTR(-EFBIG);
702 
703         ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
704         if (!ar)
705                 return ERR_PTR(-ENOMEM);
706 
707         spin_lock(&simple_transaction_lock);
708 
709         /* only one write allowed per open */
710         if (file->private_data) {
711                 spin_unlock(&simple_transaction_lock);
712                 free_page((unsigned long)ar);
713                 return ERR_PTR(-EBUSY);
714         }
715 
716         file->private_data = ar;
717 
718         spin_unlock(&simple_transaction_lock);
719 
720         if (copy_from_user(ar->data, buf, size))
721                 return ERR_PTR(-EFAULT);
722 
723         return ar->data;
724 }
725 EXPORT_SYMBOL(simple_transaction_get);
726 
727 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
728 {
729         struct simple_transaction_argresp *ar = file->private_data;
730 
731         if (!ar)
732                 return 0;
733         return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
734 }
735 EXPORT_SYMBOL(simple_transaction_read);
736 
737 int simple_transaction_release(struct inode *inode, struct file *file)
738 {
739         free_page((unsigned long)file->private_data);
740         return 0;
741 }
742 EXPORT_SYMBOL(simple_transaction_release);
743 
744 /* Simple attribute files */
745 
746 struct simple_attr {
747         int (*get)(void *, u64 *);
748         int (*set)(void *, u64);
749         char get_buf[24];       /* enough to store a u64 and "\n\0" */
750         char set_buf[24];
751         void *data;
752         const char *fmt;        /* format for read operation */
753         struct mutex mutex;     /* protects access to these buffers */
754 };
755 
756 /* simple_attr_open is called by an actual attribute open file operation
757  * to set the attribute specific access operations. */
758 int simple_attr_open(struct inode *inode, struct file *file,
759                      int (*get)(void *, u64 *), int (*set)(void *, u64),
760                      const char *fmt)
761 {
762         struct simple_attr *attr;
763 
764         attr = kmalloc(sizeof(*attr), GFP_KERNEL);
765         if (!attr)
766                 return -ENOMEM;
767 
768         attr->get = get;
769         attr->set = set;
770         attr->data = inode->i_private;
771         attr->fmt = fmt;
772         mutex_init(&attr->mutex);
773 
774         file->private_data = attr;
775 
776         return nonseekable_open(inode, file);
777 }
778 EXPORT_SYMBOL_GPL(simple_attr_open);
779 
780 int simple_attr_release(struct inode *inode, struct file *file)
781 {
782         kfree(file->private_data);
783         return 0;
784 }
785 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only?  This?  Really? */
786 
787 /* read from the buffer that is filled with the get function */
788 ssize_t simple_attr_read(struct file *file, char __user *buf,
789                          size_t len, loff_t *ppos)
790 {
791         struct simple_attr *attr;
792         size_t size;
793         ssize_t ret;
794 
795         attr = file->private_data;
796 
797         if (!attr->get)
798                 return -EACCES;
799 
800         ret = mutex_lock_interruptible(&attr->mutex);
801         if (ret)
802                 return ret;
803 
804         if (*ppos) {            /* continued read */
805                 size = strlen(attr->get_buf);
806         } else {                /* first read */
807                 u64 val;
808                 ret = attr->get(attr->data, &val);
809                 if (ret)
810                         goto out;
811 
812                 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
813                                  attr->fmt, (unsigned long long)val);
814         }
815 
816         ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
817 out:
818         mutex_unlock(&attr->mutex);
819         return ret;
820 }
821 EXPORT_SYMBOL_GPL(simple_attr_read);
822 
823 /* interpret the buffer as a number to call the set function with */
824 ssize_t simple_attr_write(struct file *file, const char __user *buf,
825                           size_t len, loff_t *ppos)
826 {
827         struct simple_attr *attr;
828         u64 val;
829         size_t size;
830         ssize_t ret;
831 
832         attr = file->private_data;
833         if (!attr->set)
834                 return -EACCES;
835 
836         ret = mutex_lock_interruptible(&attr->mutex);
837         if (ret)
838                 return ret;
839 
840         ret = -EFAULT;
841         size = min(sizeof(attr->set_buf) - 1, len);
842         if (copy_from_user(attr->set_buf, buf, size))
843                 goto out;
844 
845         attr->set_buf[size] = '\0';
846         val = simple_strtoll(attr->set_buf, NULL, 0);
847         ret = attr->set(attr->data, val);
848         if (ret == 0)
849                 ret = len; /* on success, claim we got the whole input */
850 out:
851         mutex_unlock(&attr->mutex);
852         return ret;
853 }
854 EXPORT_SYMBOL_GPL(simple_attr_write);
855 
856 /**
857  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
858  * @sb:         filesystem to do the file handle conversion on
859  * @fid:        file handle to convert
860  * @fh_len:     length of the file handle in bytes
861  * @fh_type:    type of file handle
862  * @get_inode:  filesystem callback to retrieve inode
863  *
864  * This function decodes @fid as long as it has one of the well-known
865  * Linux filehandle types and calls @get_inode on it to retrieve the
866  * inode for the object specified in the file handle.
867  */
868 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
869                 int fh_len, int fh_type, struct inode *(*get_inode)
870                         (struct super_block *sb, u64 ino, u32 gen))
871 {
872         struct inode *inode = NULL;
873 
874         if (fh_len < 2)
875                 return NULL;
876 
877         switch (fh_type) {
878         case FILEID_INO32_GEN:
879         case FILEID_INO32_GEN_PARENT:
880                 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
881                 break;
882         }
883 
884         return d_obtain_alias(inode);
885 }
886 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
887 
888 /**
889  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
890  * @sb:         filesystem to do the file handle conversion on
891  * @fid:        file handle to convert
892  * @fh_len:     length of the file handle in bytes
893  * @fh_type:    type of file handle
894  * @get_inode:  filesystem callback to retrieve inode
895  *
896  * This function decodes @fid as long as it has one of the well-known
897  * Linux filehandle types and calls @get_inode on it to retrieve the
898  * inode for the _parent_ object specified in the file handle if it
899  * is specified in the file handle, or NULL otherwise.
900  */
901 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
902                 int fh_len, int fh_type, struct inode *(*get_inode)
903                         (struct super_block *sb, u64 ino, u32 gen))
904 {
905         struct inode *inode = NULL;
906 
907         if (fh_len <= 2)
908                 return NULL;
909 
910         switch (fh_type) {
911         case FILEID_INO32_GEN_PARENT:
912                 inode = get_inode(sb, fid->i32.parent_ino,
913                                   (fh_len > 3 ? fid->i32.parent_gen : 0));
914                 break;
915         }
916 
917         return d_obtain_alias(inode);
918 }
919 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
920 
921 /**
922  * __generic_file_fsync - generic fsync implementation for simple filesystems
923  *
924  * @file:       file to synchronize
925  * @start:      start offset in bytes
926  * @end:        end offset in bytes (inclusive)
927  * @datasync:   only synchronize essential metadata if true
928  *
929  * This is a generic implementation of the fsync method for simple
930  * filesystems which track all non-inode metadata in the buffers list
931  * hanging off the address_space structure.
932  */
933 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
934                                  int datasync)
935 {
936         struct inode *inode = file->f_mapping->host;
937         int err;
938         int ret;
939 
940         err = filemap_write_and_wait_range(inode->i_mapping, start, end);
941         if (err)
942                 return err;
943 
944         inode_lock(inode);
945         ret = sync_mapping_buffers(inode->i_mapping);
946         if (!(inode->i_state & I_DIRTY_ALL))
947                 goto out;
948         if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
949                 goto out;
950 
951         err = sync_inode_metadata(inode, 1);
952         if (ret == 0)
953                 ret = err;
954 
955 out:
956         inode_unlock(inode);
957         return ret;
958 }
959 EXPORT_SYMBOL(__generic_file_fsync);
960 
961 /**
962  * generic_file_fsync - generic fsync implementation for simple filesystems
963  *                      with flush
964  * @file:       file to synchronize
965  * @start:      start offset in bytes
966  * @end:        end offset in bytes (inclusive)
967  * @datasync:   only synchronize essential metadata if true
968  *
969  */
970 
971 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
972                        int datasync)
973 {
974         struct inode *inode = file->f_mapping->host;
975         int err;
976 
977         err = __generic_file_fsync(file, start, end, datasync);
978         if (err)
979                 return err;
980         return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
981 }
982 EXPORT_SYMBOL(generic_file_fsync);
983 
984 /**
985  * generic_check_addressable - Check addressability of file system
986  * @blocksize_bits:     log of file system block size
987  * @num_blocks:         number of blocks in file system
988  *
989  * Determine whether a file system with @num_blocks blocks (and a
990  * block size of 2**@blocksize_bits) is addressable by the sector_t
991  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
992  */
993 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
994 {
995         u64 last_fs_block = num_blocks - 1;
996         u64 last_fs_page =
997                 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
998 
999         if (unlikely(num_blocks == 0))
1000                 return 0;
1001 
1002         if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1003                 return -EINVAL;
1004 
1005         if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1006             (last_fs_page > (pgoff_t)(~0ULL))) {
1007                 return -EFBIG;
1008         }
1009         return 0;
1010 }
1011 EXPORT_SYMBOL(generic_check_addressable);
1012 
1013 /*
1014  * No-op implementation of ->fsync for in-memory filesystems.
1015  */
1016 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1017 {
1018         return 0;
1019 }
1020 EXPORT_SYMBOL(noop_fsync);
1021 
1022 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1023 void kfree_link(void *p)
1024 {
1025         kfree(p);
1026 }
1027 EXPORT_SYMBOL(kfree_link);
1028 
1029 /*
1030  * nop .set_page_dirty method so that people can use .page_mkwrite on
1031  * anon inodes.
1032  */
1033 static int anon_set_page_dirty(struct page *page)
1034 {
1035         return 0;
1036 };
1037 
1038 /*
1039  * A single inode exists for all anon_inode files. Contrary to pipes,
1040  * anon_inode inodes have no associated per-instance data, so we need
1041  * only allocate one of them.
1042  */
1043 struct inode *alloc_anon_inode(struct super_block *s)
1044 {
1045         static const struct address_space_operations anon_aops = {
1046                 .set_page_dirty = anon_set_page_dirty,
1047         };
1048         struct inode *inode = new_inode_pseudo(s);
1049 
1050         if (!inode)
1051                 return ERR_PTR(-ENOMEM);
1052 
1053         inode->i_ino = get_next_ino();
1054         inode->i_mapping->a_ops = &anon_aops;
1055 
1056         /*
1057          * Mark the inode dirty from the very beginning,
1058          * that way it will never be moved to the dirty
1059          * list because mark_inode_dirty() will think
1060          * that it already _is_ on the dirty list.
1061          */
1062         inode->i_state = I_DIRTY;
1063         inode->i_mode = S_IRUSR | S_IWUSR;
1064         inode->i_uid = current_fsuid();
1065         inode->i_gid = current_fsgid();
1066         inode->i_flags |= S_PRIVATE;
1067         inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1068         return inode;
1069 }
1070 EXPORT_SYMBOL(alloc_anon_inode);
1071 
1072 /**
1073  * simple_nosetlease - generic helper for prohibiting leases
1074  * @filp: file pointer
1075  * @arg: type of lease to obtain
1076  * @flp: new lease supplied for insertion
1077  * @priv: private data for lm_setup operation
1078  *
1079  * Generic helper for filesystems that do not wish to allow leases to be set.
1080  * All arguments are ignored and it just returns -EINVAL.
1081  */
1082 int
1083 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1084                   void **priv)
1085 {
1086         return -EINVAL;
1087 }
1088 EXPORT_SYMBOL(simple_nosetlease);
1089 
1090 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1091                             struct delayed_call *done)
1092 {
1093         return inode->i_link;
1094 }
1095 EXPORT_SYMBOL(simple_get_link);
1096 
1097 const struct inode_operations simple_symlink_inode_operations = {
1098         .get_link = simple_get_link,
1099         .readlink = generic_readlink
1100 };
1101 EXPORT_SYMBOL(simple_symlink_inode_operations);
1102 
1103 /*
1104  * Operations for a permanently empty directory.
1105  */
1106 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1107 {
1108         return ERR_PTR(-ENOENT);
1109 }
1110 
1111 static int empty_dir_getattr(struct vfsmount *mnt, struct dentry *dentry,
1112                                  struct kstat *stat)
1113 {
1114         struct inode *inode = d_inode(dentry);
1115         generic_fillattr(inode, stat);
1116         return 0;
1117 }
1118 
1119 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1120 {
1121         return -EPERM;
1122 }
1123 
1124 static int empty_dir_setxattr(struct dentry *dentry, const char *name,
1125                               const void *value, size_t size, int flags)
1126 {
1127         return -EOPNOTSUPP;
1128 }
1129 
1130 static ssize_t empty_dir_getxattr(struct dentry *dentry, const char *name,
1131                                   void *value, size_t size)
1132 {
1133         return -EOPNOTSUPP;
1134 }
1135 
1136 static int empty_dir_removexattr(struct dentry *dentry, const char *name)
1137 {
1138         return -EOPNOTSUPP;
1139 }
1140 
1141 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1142 {
1143         return -EOPNOTSUPP;
1144 }
1145 
1146 static const struct inode_operations empty_dir_inode_operations = {
1147         .lookup         = empty_dir_lookup,
1148         .permission     = generic_permission,
1149         .setattr        = empty_dir_setattr,
1150         .getattr        = empty_dir_getattr,
1151         .setxattr       = empty_dir_setxattr,
1152         .getxattr       = empty_dir_getxattr,
1153         .removexattr    = empty_dir_removexattr,
1154         .listxattr      = empty_dir_listxattr,
1155 };
1156 
1157 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1158 {
1159         /* An empty directory has two entries . and .. at offsets 0 and 1 */
1160         return generic_file_llseek_size(file, offset, whence, 2, 2);
1161 }
1162 
1163 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1164 {
1165         dir_emit_dots(file, ctx);
1166         return 0;
1167 }
1168 
1169 static const struct file_operations empty_dir_operations = {
1170         .llseek         = empty_dir_llseek,
1171         .read           = generic_read_dir,
1172         .iterate        = empty_dir_readdir,
1173         .fsync          = noop_fsync,
1174 };
1175 
1176 
1177 void make_empty_dir_inode(struct inode *inode)
1178 {
1179         set_nlink(inode, 2);
1180         inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1181         inode->i_uid = GLOBAL_ROOT_UID;
1182         inode->i_gid = GLOBAL_ROOT_GID;
1183         inode->i_rdev = 0;
1184         inode->i_size = 0;
1185         inode->i_blkbits = PAGE_SHIFT;
1186         inode->i_blocks = 0;
1187 
1188         inode->i_op = &empty_dir_inode_operations;
1189         inode->i_fop = &empty_dir_operations;
1190 }
1191 
1192 bool is_empty_dir_inode(struct inode *inode)
1193 {
1194         return (inode->i_fop == &empty_dir_operations) &&
1195                 (inode->i_op == &empty_dir_inode_operations);
1196 }
1197