linux/fs/libfs.c
<<
>>
Prefs
   1/*
   2 *      fs/libfs.c
   3 *      Library for filesystems writers.
   4 */
   5
   6#include <linux/export.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> /* sync_mapping_buffers */
  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, unsigned int flags)
  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 = QSTR_INIT(".", 1);
  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 whence)
  85{
  86        struct dentry *dentry = file->f_path.dentry;
  87        mutex_lock(&dentry->d_inode->i_mutex);
  88        switch (whence) {
  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;
 226        struct dentry *dentry;
 227        struct inode *root;
 228        struct qstr d_name = QSTR_INIT(name, strlen(name));
 229
 230        s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
 231        if (IS_ERR(s))
 232                return ERR_CAST(s);
 233
 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_open(struct inode *inode, struct file *file)
 268{
 269        if (inode->i_private)
 270                file->private_data = inode->i_private;
 271        return 0;
 272}
 273
 274int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
 275{
 276        struct inode *inode = old_dentry->d_inode;
 277
 278        inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
 279        inc_nlink(inode);
 280        ihold(inode);
 281        dget(dentry);
 282        d_instantiate(dentry, inode);
 283        return 0;
 284}
 285
 286int simple_empty(struct dentry *dentry)
 287{
 288        struct dentry *child;
 289        int ret = 0;
 290
 291        spin_lock(&dentry->d_lock);
 292        list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
 293                spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
 294                if (simple_positive(child)) {
 295                        spin_unlock(&child->d_lock);
 296                        goto out;
 297                }
 298                spin_unlock(&child->d_lock);
 299        }
 300        ret = 1;
 301out:
 302        spin_unlock(&dentry->d_lock);
 303        return ret;
 304}
 305
 306int simple_unlink(struct inode *dir, struct dentry *dentry)
 307{
 308        struct inode *inode = dentry->d_inode;
 309
 310        inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
 311        drop_nlink(inode);
 312        dput(dentry);
 313        return 0;
 314}
 315
 316int simple_rmdir(struct inode *dir, struct dentry *dentry)
 317{
 318        if (!simple_empty(dentry))
 319                return -ENOTEMPTY;
 320
 321        drop_nlink(dentry->d_inode);
 322        simple_unlink(dir, dentry);
 323        drop_nlink(dir);
 324        return 0;
 325}
 326
 327int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
 328                struct inode *new_dir, struct dentry *new_dentry)
 329{
 330        struct inode *inode = old_dentry->d_inode;
 331        int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
 332
 333        if (!simple_empty(new_dentry))
 334                return -ENOTEMPTY;
 335
 336        if (new_dentry->d_inode) {
 337                simple_unlink(new_dir, new_dentry);
 338                if (they_are_dirs) {
 339                        drop_nlink(new_dentry->d_inode);
 340                        drop_nlink(old_dir);
 341                }
 342        } else if (they_are_dirs) {
 343                drop_nlink(old_dir);
 344                inc_nlink(new_dir);
 345        }
 346
 347        old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
 348                new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
 349
 350        return 0;
 351}
 352
 353/**
 354 * simple_setattr - setattr for simple filesystem
 355 * @dentry: dentry
 356 * @iattr: iattr structure
 357 *
 358 * Returns 0 on success, -error on failure.
 359 *
 360 * simple_setattr is a simple ->setattr implementation without a proper
 361 * implementation of size changes.
 362 *
 363 * It can either be used for in-memory filesystems or special files
 364 * on simple regular filesystems.  Anything that needs to change on-disk
 365 * or wire state on size changes needs its own setattr method.
 366 */
 367int simple_setattr(struct dentry *dentry, struct iattr *iattr)
 368{
 369        struct inode *inode = dentry->d_inode;
 370        int error;
 371
 372        error = inode_change_ok(inode, iattr);
 373        if (error)
 374                return error;
 375
 376        if (iattr->ia_valid & ATTR_SIZE)
 377                truncate_setsize(inode, iattr->ia_size);
 378        setattr_copy(inode, iattr);
 379        mark_inode_dirty(inode);
 380        return 0;
 381}
 382EXPORT_SYMBOL(simple_setattr);
 383
 384int simple_readpage(struct file *file, struct page *page)
 385{
 386        clear_highpage(page);
 387        flush_dcache_page(page);
 388        SetPageUptodate(page);
 389        unlock_page(page);
 390        return 0;
 391}
 392
 393int simple_write_begin(struct file *file, struct address_space *mapping,
 394                        loff_t pos, unsigned len, unsigned flags,
 395                        struct page **pagep, void **fsdata)
 396{
 397        struct page *page;
 398        pgoff_t index;
 399
 400        index = pos >> PAGE_CACHE_SHIFT;
 401
 402        page = grab_cache_page_write_begin(mapping, index, flags);
 403        if (!page)
 404                return -ENOMEM;
 405
 406        *pagep = page;
 407
 408        if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
 409                unsigned from = pos & (PAGE_CACHE_SIZE - 1);
 410
 411                zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
 412        }
 413        return 0;
 414}
 415
 416/**
 417 * simple_write_end - .write_end helper for non-block-device FSes
 418 * @available: See .write_end of address_space_operations
 419 * @file:               "
 420 * @mapping:            "
 421 * @pos:                "
 422 * @len:                "
 423 * @copied:             "
 424 * @page:               "
 425 * @fsdata:             "
 426 *
 427 * simple_write_end does the minimum needed for updating a page after writing is
 428 * done. It has the same API signature as the .write_end of
 429 * address_space_operations vector. So it can just be set onto .write_end for
 430 * FSes that don't need any other processing. i_mutex is assumed to be held.
 431 * Block based filesystems should use generic_write_end().
 432 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
 433 * is not called, so a filesystem that actually does store data in .write_inode
 434 * should extend on what's done here with a call to mark_inode_dirty() in the
 435 * case that i_size has changed.
 436 */
 437int simple_write_end(struct file *file, struct address_space *mapping,
 438                        loff_t pos, unsigned len, unsigned copied,
 439                        struct page *page, void *fsdata)
 440{
 441        struct inode *inode = page->mapping->host;
 442        loff_t last_pos = pos + copied;
 443
 444        /* zero the stale part of the page if we did a short copy */
 445        if (copied < len) {
 446                unsigned from = pos & (PAGE_CACHE_SIZE - 1);
 447
 448                zero_user(page, from + copied, len - copied);
 449        }
 450
 451        if (!PageUptodate(page))
 452                SetPageUptodate(page);
 453        /*
 454         * No need to use i_size_read() here, the i_size
 455         * cannot change under us because we hold the i_mutex.
 456         */
 457        if (last_pos > inode->i_size)
 458                i_size_write(inode, last_pos);
 459
 460        set_page_dirty(page);
 461        unlock_page(page);
 462        page_cache_release(page);
 463
 464        return copied;
 465}
 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 */
 472int 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_CACHE_SIZE;
 481        s->s_blocksize_bits = PAGE_CACHE_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;
 528out:
 529        d_genocide(root);
 530        shrink_dcache_parent(root);
 531        dput(root);
 532        return -ENOMEM;
 533}
 534
 535static DEFINE_SPINLOCK(pin_fs_lock);
 536
 537int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
 538{
 539        struct vfsmount *mnt = NULL;
 540        spin_lock(&pin_fs_lock);
 541        if (unlikely(!*mount)) {
 542                spin_unlock(&pin_fs_lock);
 543                mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
 544                if (IS_ERR(mnt))
 545                        return PTR_ERR(mnt);
 546                spin_lock(&pin_fs_lock);
 547                if (!*mount)
 548                        *mount = mnt;
 549        }
 550        mntget(*mount);
 551        ++*count;
 552        spin_unlock(&pin_fs_lock);
 553        mntput(mnt);
 554        return 0;
 555}
 556
 557void simple_release_fs(struct vfsmount **mount, int *count)
 558{
 559        struct vfsmount *mnt;
 560        spin_lock(&pin_fs_lock);
 561        mnt = *mount;
 562        if (!--*count)
 563                *mount = NULL;
 564        spin_unlock(&pin_fs_lock);
 565        mntput(mnt);
 566}
 567
 568/**
 569 * simple_read_from_buffer - copy data from the buffer to user space
 570 * @to: the user space buffer to read to
 571 * @count: the maximum number of bytes to read
 572 * @ppos: the current position in the buffer
 573 * @from: the buffer to read from
 574 * @available: the size of the buffer
 575 *
 576 * The simple_read_from_buffer() function reads up to @count bytes from the
 577 * buffer @from at offset @ppos into the user space address starting at @to.
 578 *
 579 * On success, the number of bytes read is returned and the offset @ppos is
 580 * advanced by this number, or negative value is returned on error.
 581 **/
 582ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
 583                                const void *from, size_t available)
 584{
 585        loff_t pos = *ppos;
 586        size_t ret;
 587
 588        if (pos < 0)
 589                return -EINVAL;
 590        if (pos >= available || !count)
 591                return 0;
 592        if (count > available - pos)
 593                count = available - pos;
 594        ret = copy_to_user(to, from + pos, count);
 595        if (ret == count)
 596                return -EFAULT;
 597        count -= ret;
 598        *ppos = pos + count;
 599        return count;
 600}
 601
 602/**
 603 * simple_write_to_buffer - copy data from user space to the buffer
 604 * @to: the buffer to write to
 605 * @available: the size of the buffer
 606 * @ppos: the current position in the buffer
 607 * @from: the user space buffer to read from
 608 * @count: the maximum number of bytes to read
 609 *
 610 * The simple_write_to_buffer() function reads up to @count bytes from the user
 611 * space address starting at @from into the buffer @to at offset @ppos.
 612 *
 613 * On success, the number of bytes written is returned and the offset @ppos is
 614 * advanced by this number, or negative value is returned on error.
 615 **/
 616ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
 617                const void __user *from, size_t count)
 618{
 619        loff_t pos = *ppos;
 620        size_t res;
 621
 622        if (pos < 0)
 623                return -EINVAL;
 624        if (pos >= available || !count)
 625                return 0;
 626        if (count > available - pos)
 627                count = available - pos;
 628        res = copy_from_user(to + pos, from, count);
 629        if (res == count)
 630                return -EFAULT;
 631        count -= res;
 632        *ppos = pos + count;
 633        return count;
 634}
 635
 636/**
 637 * memory_read_from_buffer - copy data from the buffer
 638 * @to: the kernel space buffer to read to
 639 * @count: the maximum number of bytes to read
 640 * @ppos: the current position in the buffer
 641 * @from: the buffer to read from
 642 * @available: the size of the buffer
 643 *
 644 * The memory_read_from_buffer() function reads up to @count bytes from the
 645 * buffer @from at offset @ppos into the kernel space address starting at @to.
 646 *
 647 * On success, the number of bytes read is returned and the offset @ppos is
 648 * advanced by this number, or negative value is returned on error.
 649 **/
 650ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
 651                                const void *from, size_t available)
 652{
 653        loff_t pos = *ppos;
 654
 655        if (pos < 0)
 656                return -EINVAL;
 657        if (pos >= available)
 658                return 0;
 659        if (count > available - pos)
 660                count = available - pos;
 661        memcpy(to, from + pos, count);
 662        *ppos = pos + count;
 663
 664        return count;
 665}
 666
 667/*
 668 * Transaction based IO.
 669 * The file expects a single write which triggers the transaction, and then
 670 * possibly a read which collects the result - which is stored in a
 671 * file-local buffer.
 672 */
 673
 674void simple_transaction_set(struct file *file, size_t n)
 675{
 676        struct simple_transaction_argresp *ar = file->private_data;
 677
 678        BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
 679
 680        /*
 681         * The barrier ensures that ar->size will really remain zero until
 682         * ar->data is ready for reading.
 683         */
 684        smp_mb();
 685        ar->size = n;
 686}
 687
 688char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
 689{
 690        struct simple_transaction_argresp *ar;
 691        static DEFINE_SPINLOCK(simple_transaction_lock);
 692
 693        if (size > SIMPLE_TRANSACTION_LIMIT - 1)
 694                return ERR_PTR(-EFBIG);
 695
 696        ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
 697        if (!ar)
 698                return ERR_PTR(-ENOMEM);
 699
 700        spin_lock(&simple_transaction_lock);
 701
 702        /* only one write allowed per open */
 703        if (file->private_data) {
 704                spin_unlock(&simple_transaction_lock);
 705                free_page((unsigned long)ar);
 706                return ERR_PTR(-EBUSY);
 707        }
 708
 709        file->private_data = ar;
 710
 711        spin_unlock(&simple_transaction_lock);
 712
 713        if (copy_from_user(ar->data, buf, size))
 714                return ERR_PTR(-EFAULT);
 715
 716        return ar->data;
 717}
 718
 719ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 720{
 721        struct simple_transaction_argresp *ar = file->private_data;
 722
 723        if (!ar)
 724                return 0;
 725        return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
 726}
 727
 728int simple_transaction_release(struct inode *inode, struct file *file)
 729{
 730        free_page((unsigned long)file->private_data);
 731        return 0;
 732}
 733
 734/* Simple attribute files */
 735
 736struct simple_attr {
 737        int (*get)(void *, u64 *);
 738        int (*set)(void *, u64);
 739        char get_buf[24];       /* enough to store a u64 and "\n\0" */
 740        char set_buf[24];
 741        void *data;
 742        const char *fmt;        /* format for read operation */
 743        struct mutex mutex;     /* protects access to these buffers */
 744};
 745
 746/* simple_attr_open is called by an actual attribute open file operation
 747 * to set the attribute specific access operations. */
 748int simple_attr_open(struct inode *inode, struct file *file,
 749                     int (*get)(void *, u64 *), int (*set)(void *, u64),
 750                     const char *fmt)
 751{
 752        struct simple_attr *attr;
 753
 754        attr = kmalloc(sizeof(*attr), GFP_KERNEL);
 755        if (!attr)
 756                return -ENOMEM;
 757
 758        attr->get = get;
 759        attr->set = set;
 760        attr->data = inode->i_private;
 761        attr->fmt = fmt;
 762        mutex_init(&attr->mutex);
 763
 764        file->private_data = attr;
 765
 766        return nonseekable_open(inode, file);
 767}
 768
 769int simple_attr_release(struct inode *inode, struct file *file)
 770{
 771        kfree(file->private_data);
 772        return 0;
 773}
 774
 775/* read from the buffer that is filled with the get function */
 776ssize_t simple_attr_read(struct file *file, char __user *buf,
 777                         size_t len, loff_t *ppos)
 778{
 779        struct simple_attr *attr;
 780        size_t size;
 781        ssize_t ret;
 782
 783        attr = file->private_data;
 784
 785        if (!attr->get)
 786                return -EACCES;
 787
 788        ret = mutex_lock_interruptible(&attr->mutex);
 789        if (ret)
 790                return ret;
 791
 792        if (*ppos) {            /* continued read */
 793                size = strlen(attr->get_buf);
 794        } else {                /* first read */
 795                u64 val;
 796                ret = attr->get(attr->data, &val);
 797                if (ret)
 798                        goto out;
 799
 800                size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
 801                                 attr->fmt, (unsigned long long)val);
 802        }
 803
 804        ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
 805out:
 806        mutex_unlock(&attr->mutex);
 807        return ret;
 808}
 809
 810/* interpret the buffer as a number to call the set function with */
 811ssize_t simple_attr_write(struct file *file, const char __user *buf,
 812                          size_t len, loff_t *ppos)
 813{
 814        struct simple_attr *attr;
 815        u64 val;
 816        size_t size;
 817        ssize_t ret;
 818
 819        attr = file->private_data;
 820        if (!attr->set)
 821                return -EACCES;
 822
 823        ret = mutex_lock_interruptible(&attr->mutex);
 824        if (ret)
 825                return ret;
 826
 827        ret = -EFAULT;
 828        size = min(sizeof(attr->set_buf) - 1, len);
 829        if (copy_from_user(attr->set_buf, buf, size))
 830                goto out;
 831
 832        attr->set_buf[size] = '\0';
 833        val = simple_strtoll(attr->set_buf, NULL, 0);
 834        ret = attr->set(attr->data, val);
 835        if (ret == 0)
 836                ret = len; /* on success, claim we got the whole input */
 837out:
 838        mutex_unlock(&attr->mutex);
 839        return ret;
 840}
 841
 842/**
 843 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
 844 * @sb:         filesystem to do the file handle conversion on
 845 * @fid:        file handle to convert
 846 * @fh_len:     length of the file handle in bytes
 847 * @fh_type:    type of file handle
 848 * @get_inode:  filesystem callback to retrieve inode
 849 *
 850 * This function decodes @fid as long as it has one of the well-known
 851 * Linux filehandle types and calls @get_inode on it to retrieve the
 852 * inode for the object specified in the file handle.
 853 */
 854struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
 855                int fh_len, int fh_type, struct inode *(*get_inode)
 856                        (struct super_block *sb, u64 ino, u32 gen))
 857{
 858        struct inode *inode = NULL;
 859
 860        if (fh_len < 2)
 861                return NULL;
 862
 863        switch (fh_type) {
 864        case FILEID_INO32_GEN:
 865        case FILEID_INO32_GEN_PARENT:
 866                inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
 867                break;
 868        }
 869
 870        return d_obtain_alias(inode);
 871}
 872EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
 873
 874/**
 875 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
 876 * @sb:         filesystem to do the file handle conversion on
 877 * @fid:        file handle to convert
 878 * @fh_len:     length of the file handle in bytes
 879 * @fh_type:    type of file handle
 880 * @get_inode:  filesystem callback to retrieve inode
 881 *
 882 * This function decodes @fid as long as it has one of the well-known
 883 * Linux filehandle types and calls @get_inode on it to retrieve the
 884 * inode for the _parent_ object specified in the file handle if it
 885 * is specified in the file handle, or NULL otherwise.
 886 */
 887struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
 888                int fh_len, int fh_type, struct inode *(*get_inode)
 889                        (struct super_block *sb, u64 ino, u32 gen))
 890{
 891        struct inode *inode = NULL;
 892
 893        if (fh_len <= 2)
 894                return NULL;
 895
 896        switch (fh_type) {
 897        case FILEID_INO32_GEN_PARENT:
 898                inode = get_inode(sb, fid->i32.parent_ino,
 899                                  (fh_len > 3 ? fid->i32.parent_gen : 0));
 900                break;
 901        }
 902
 903        return d_obtain_alias(inode);
 904}
 905EXPORT_SYMBOL_GPL(generic_fh_to_parent);
 906
 907/**
 908 * generic_file_fsync - generic fsync implementation for simple filesystems
 909 * @file:       file to synchronize
 910 * @datasync:   only synchronize essential metadata if true
 911 *
 912 * This is a generic implementation of the fsync method for simple
 913 * filesystems which track all non-inode metadata in the buffers list
 914 * hanging off the address_space structure.
 915 */
 916int generic_file_fsync(struct file *file, loff_t start, loff_t end,
 917                       int datasync)
 918{
 919        struct inode *inode = file->f_mapping->host;
 920        int err;
 921        int ret;
 922
 923        err = filemap_write_and_wait_range(inode->i_mapping, start, end);
 924        if (err)
 925                return err;
 926
 927        mutex_lock(&inode->i_mutex);
 928        ret = sync_mapping_buffers(inode->i_mapping);
 929        if (!(inode->i_state & I_DIRTY))
 930                goto out;
 931        if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
 932                goto out;
 933
 934        err = sync_inode_metadata(inode, 1);
 935        if (ret == 0)
 936                ret = err;
 937out:
 938        mutex_unlock(&inode->i_mutex);
 939        return ret;
 940}
 941EXPORT_SYMBOL(generic_file_fsync);
 942
 943/**
 944 * generic_check_addressable - Check addressability of file system
 945 * @blocksize_bits:     log of file system block size
 946 * @num_blocks:         number of blocks in file system
 947 *
 948 * Determine whether a file system with @num_blocks blocks (and a
 949 * block size of 2**@blocksize_bits) is addressable by the sector_t
 950 * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
 951 */
 952int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
 953{
 954        u64 last_fs_block = num_blocks - 1;
 955        u64 last_fs_page =
 956                last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
 957
 958        if (unlikely(num_blocks == 0))
 959                return 0;
 960
 961        if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
 962                return -EINVAL;
 963
 964        if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
 965            (last_fs_page > (pgoff_t)(~0ULL))) {
 966                return -EFBIG;
 967        }
 968        return 0;
 969}
 970EXPORT_SYMBOL(generic_check_addressable);
 971
 972/*
 973 * No-op implementation of ->fsync for in-memory filesystems.
 974 */
 975int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
 976{
 977        return 0;
 978}
 979
 980EXPORT_SYMBOL(dcache_dir_close);
 981EXPORT_SYMBOL(dcache_dir_lseek);
 982EXPORT_SYMBOL(dcache_dir_open);
 983EXPORT_SYMBOL(dcache_readdir);
 984EXPORT_SYMBOL(generic_read_dir);
 985EXPORT_SYMBOL(mount_pseudo);
 986EXPORT_SYMBOL(simple_write_begin);
 987EXPORT_SYMBOL(simple_write_end);
 988EXPORT_SYMBOL(simple_dir_inode_operations);
 989EXPORT_SYMBOL(simple_dir_operations);
 990EXPORT_SYMBOL(simple_empty);
 991EXPORT_SYMBOL(simple_fill_super);
 992EXPORT_SYMBOL(simple_getattr);
 993EXPORT_SYMBOL(simple_open);
 994EXPORT_SYMBOL(simple_link);
 995EXPORT_SYMBOL(simple_lookup);
 996EXPORT_SYMBOL(simple_pin_fs);
 997EXPORT_SYMBOL(simple_readpage);
 998EXPORT_SYMBOL(simple_release_fs);
 999EXPORT_SYMBOL(simple_rename);
1000EXPORT_SYMBOL(simple_rmdir);
1001EXPORT_SYMBOL(simple_statfs);
1002EXPORT_SYMBOL(noop_fsync);
1003EXPORT_SYMBOL(simple_unlink);
1004EXPORT_SYMBOL(simple_read_from_buffer);
1005EXPORT_SYMBOL(simple_write_to_buffer);
1006EXPORT_SYMBOL(memory_read_from_buffer);
1007EXPORT_SYMBOL(simple_transaction_set);
1008EXPORT_SYMBOL(simple_transaction_get);
1009EXPORT_SYMBOL(simple_transaction_read);
1010EXPORT_SYMBOL(simple_transaction_release);
1011EXPORT_SYMBOL_GPL(simple_attr_open);
1012EXPORT_SYMBOL_GPL(simple_attr_release);
1013EXPORT_SYMBOL_GPL(simple_attr_read);
1014EXPORT_SYMBOL_GPL(simple_attr_write);
1015