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