SAFE public projects git trees. - safe/jmp/linux-2.6/blob - fs/btrfs/super.c

   1 #include <linux/module.h>
   2 #include <linux/buffer_head.h>
   3 #include <linux/fs.h>
   4 #include <linux/pagemap.h>
   5 #include <linux/highmem.h>
   6 #include <linux/time.h>
   7 #include <linux/init.h>
   8 #include <linux/string.h>
   9 #include <linux/smp_lock.h>
  10 #include <linux/backing-dev.h>
  11 #include <linux/mpage.h>
  12 #include <linux/swap.h>
  13 #include <linux/writeback.h>
  14 #include "ctree.h"
  15 #include "disk-io.h"
  16 #include "transaction.h"
  17 #include "btrfs_inode.h"
  18 #include "ioctl.h"
  19
  20 void btrfs_fsinfo_release(struct kobject *obj)
  21 {
  22         struct btrfs_fs_info *fsinfo = container_of(obj,
  23                                             struct btrfs_fs_info, kobj);
  24         kfree(fsinfo);
  25 }
  26
  27 struct kobj_type btrfs_fsinfo_ktype = {
  28         .release = btrfs_fsinfo_release,
  29 };
  30
  31 struct btrfs_iget_args {
  32         u64 ino;
  33         struct btrfs_root *root;
  34 };
  35
  36 decl_subsys(btrfs, &btrfs_fsinfo_ktype, NULL);
  37
  38 #define BTRFS_SUPER_MAGIC 0x9123682E
  39
  40 static struct inode_operations btrfs_dir_inode_operations;
  41 static struct inode_operations btrfs_dir_ro_inode_operations;
  42 static struct super_operations btrfs_super_ops;
  43 static struct file_operations btrfs_dir_file_operations;
  44 static struct inode_operations btrfs_file_inode_operations;
  45 static struct address_space_operations btrfs_aops;
  46 static struct file_operations btrfs_file_operations;
  47
  48 static void btrfs_read_locked_inode(struct inode *inode)
  49 {
  50         struct btrfs_path *path;
  51         struct btrfs_inode_item *inode_item;
  52         struct btrfs_root *root = BTRFS_I(inode)->root;
  53         struct btrfs_key location;
  54         int ret;
  55
  56         path = btrfs_alloc_path();
  57         BUG_ON(!path);
  58         btrfs_init_path(path);
  59         mutex_lock(&root->fs_info->fs_mutex);
  60
  61         memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
  62         ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
  63         if (ret) {
  64                 btrfs_free_path(path);
  65                 goto make_bad;
  66         }
  67         inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
  68                                   path->slots[0],
  69                                   struct btrfs_inode_item);
  70
  71         inode->i_mode = btrfs_inode_mode(inode_item);
  72         inode->i_nlink = btrfs_inode_nlink(inode_item);
  73         inode->i_uid = btrfs_inode_uid(inode_item);
  74         inode->i_gid = btrfs_inode_gid(inode_item);
  75         inode->i_size = btrfs_inode_size(inode_item);
  76         inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
  77         inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
  78         inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
  79         inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
  80         inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
  81         inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
  82         inode->i_blocks = btrfs_inode_nblocks(inode_item);
  83         inode->i_generation = btrfs_inode_generation(inode_item);
  84
  85         btrfs_free_path(path);
  86         inode_item = NULL;
  87
  88         mutex_unlock(&root->fs_info->fs_mutex);
  89
  90         switch (inode->i_mode & S_IFMT) {
  91 #if 0
  92         default:
  93                 init_special_inode(inode, inode->i_mode,
  94                                    btrfs_inode_rdev(inode_item));
  95                 break;
  96 #endif
  97         case S_IFREG:
  98                 inode->i_mapping->a_ops = &btrfs_aops;
  99                 inode->i_fop = &btrfs_file_operations;
 100                 inode->i_op = &btrfs_file_inode_operations;
 101                 break;
 102         case S_IFDIR:
 103                 inode->i_fop = &btrfs_dir_file_operations;
 104                 if (root == root->fs_info->tree_root)
 105                         inode->i_op = &btrfs_dir_ro_inode_operations;
 106                 else
 107                         inode->i_op = &btrfs_dir_inode_operations;
 108                 break;
 109         case S_IFLNK:
 110                 // inode->i_op = &page_symlink_inode_operations;
 111                 break;
 112         }
 113         return;
 114
 115 make_bad:
 116         btrfs_release_path(root, path);
 117         btrfs_free_path(path);
 118         mutex_unlock(&root->fs_info->fs_mutex);
 119         make_bad_inode(inode);
 120 }
 121
 122 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 123                               struct btrfs_root *root,
 124                               struct inode *dir,
 125                               struct dentry *dentry)
 126 {
 127         struct btrfs_path *path;
 128         const char *name = dentry->d_name.name;
 129         int name_len = dentry->d_name.len;
 130         int ret;
 131         u64 objectid;
 132         struct btrfs_dir_item *di;
 133
 134         path = btrfs_alloc_path();
 135         BUG_ON(!path);
 136         btrfs_init_path(path);
 137         ret = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
 138                                     name, name_len, -1);
 139         if (ret < 0)
 140                 goto err;
 141         if (ret > 0) {
 142                 ret = -ENOENT;
 143                 goto err;
 144         }
 145         di = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
 146                             struct btrfs_dir_item);
 147         objectid = btrfs_disk_key_objectid(&di->location);
 148
 149         ret = btrfs_del_item(trans, root, path);
 150         BUG_ON(ret);
 151
 152         btrfs_release_path(root, path);
 153         ret = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
 154                                           objectid, -1);
 155         BUG_ON(ret);
 156         ret = btrfs_del_item(trans, root, path);
 157         BUG_ON(ret);
 158         dentry->d_inode->i_ctime = dir->i_ctime;
 159 err:
 160         btrfs_release_path(root, path);
 161         btrfs_free_path(path);
 162         if (ret == 0) {
 163                 inode_dec_link_count(dentry->d_inode);
 164                 dir->i_size -= name_len * 2;
 165                 mark_inode_dirty(dir);
 166         }
 167         return ret;
 168 }
 169
 170 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 171 {
 172         struct btrfs_root *root;
 173         struct btrfs_trans_handle *trans;
 174         int ret;
 175
 176         root = BTRFS_I(dir)->root;
 177         mutex_lock(&root->fs_info->fs_mutex);
 178         trans = btrfs_start_transaction(root, 1);
 179         ret = btrfs_unlink_trans(trans, root, dir, dentry);
 180         btrfs_end_transaction(trans, root);
 181         mutex_unlock(&root->fs_info->fs_mutex);
 182         return ret;
 183 }
 184
 185 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 186 {
 187         struct inode *inode = dentry->d_inode;
 188         int err;
 189         int ret;
 190         struct btrfs_root *root = BTRFS_I(dir)->root;
 191         struct btrfs_path *path;
 192         struct btrfs_key key;
 193         struct btrfs_trans_handle *trans;
 194         struct btrfs_key found_key;
 195         int found_type;
 196         struct btrfs_leaf *leaf;
 197         char *goodnames = "..";
 198
 199         path = btrfs_alloc_path();
 200         BUG_ON(!path);
 201         btrfs_init_path(path);
 202         mutex_lock(&root->fs_info->fs_mutex);
 203         trans = btrfs_start_transaction(root, 1);
 204         key.objectid = inode->i_ino;
 205         key.offset = (u64)-1;
 206         key.flags = (u32)-1;
 207         while(1) {
 208                 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 209                 if (ret < 0) {
 210                         err = ret;
 211                         goto out;
 212                 }
 213                 BUG_ON(ret == 0);
 214                 if (path->slots[0] == 0) {
 215                         err = -ENOENT;
 216                         goto out;
 217                 }
 218                 path->slots[0]--;
 219                 leaf = btrfs_buffer_leaf(path->nodes[0]);
 220                 btrfs_disk_key_to_cpu(&found_key,
 221                                       &leaf->items[path->slots[0]].key);
 222                 found_type = btrfs_key_type(&found_key);
 223                 if (found_key.objectid != inode->i_ino) {
 224                         err = -ENOENT;
 225                         goto out;
 226                 }
 227                 if ((found_type != BTRFS_DIR_ITEM_KEY &&
 228                      found_type != BTRFS_DIR_INDEX_KEY) ||
 229                     (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
 230                     !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
 231                         err = -ENOTEMPTY;
 232                         goto out;
 233                 }
 234                 ret = btrfs_del_item(trans, root, path);
 235                 BUG_ON(ret);
 236
 237                 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
 238                         break;
 239                 btrfs_release_path(root, path);
 240         }
 241         ret = 0;
 242         btrfs_release_path(root, path);
 243
 244         /* now the directory is empty */
 245         err = btrfs_unlink_trans(trans, root, dir, dentry);
 246         if (!err) {
 247                 inode->i_size = 0;
 248         }
 249 out:
 250         btrfs_release_path(root, path);
 251         btrfs_free_path(path);
 252         mutex_unlock(&root->fs_info->fs_mutex);
 253         ret = btrfs_end_transaction(trans, root);
 254         if (ret && !err)
 255                 err = ret;
 256         return err;
 257 }
 258
 259 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
 260                             struct btrfs_root *root,
 261                             struct inode *inode)
 262 {
 263         struct btrfs_path *path;
 264         int ret;
 265
 266         clear_inode(inode);
 267
 268         path = btrfs_alloc_path();
 269         BUG_ON(!path);
 270         btrfs_init_path(path);
 271         ret = btrfs_lookup_inode(trans, root, path,
 272                                  &BTRFS_I(inode)->location, -1);
 273         BUG_ON(ret);
 274         ret = btrfs_del_item(trans, root, path);
 275         BUG_ON(ret);
 276         btrfs_free_path(path);
 277         return ret;
 278 }
 279
 280 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 281                                    struct btrfs_root *root,
 282                                    struct inode *inode)
 283 {
 284         int ret;
 285         struct btrfs_path *path;
 286         struct btrfs_key key;
 287         struct btrfs_disk_key *found_key;
 288         struct btrfs_leaf *leaf;
 289         struct btrfs_file_extent_item *fi = NULL;
 290         u64 extent_start = 0;
 291         u64 extent_num_blocks = 0;
 292         int found_extent;
 293
 294         path = btrfs_alloc_path();
 295         BUG_ON(!path);
 296         /* FIXME, add redo link to tree so we don't leak on crash */
 297         key.objectid = inode->i_ino;
 298         key.offset = (u64)-1;
 299         key.flags = 0;
 300         /*
 301          * use BTRFS_CSUM_ITEM_KEY because it is larger than inline keys
 302          * or extent data
 303          */
 304         btrfs_set_key_type(&key, BTRFS_CSUM_ITEM_KEY);
 305         while(1) {
 306                 btrfs_init_path(path);
 307                 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 308                 if (ret < 0) {
 309                         goto error;
 310                 }
 311                 if (ret > 0) {
 312                         BUG_ON(path->slots[0] == 0);
 313                         path->slots[0]--;
 314                 }
 315                 leaf = btrfs_buffer_leaf(path->nodes[0]);
 316                 found_key = &leaf->items[path->slots[0]].key;
 317                 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
 318                         break;
 319                 if (btrfs_disk_key_type(found_key) != BTRFS_CSUM_ITEM_KEY &&
 320                     btrfs_disk_key_type(found_key) != BTRFS_INLINE_DATA_KEY &&
 321                     btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY)
 322                         break;
 323                 if (btrfs_disk_key_offset(found_key) < inode->i_size)
 324                         break;
 325                 if (btrfs_disk_key_type(found_key) == BTRFS_EXTENT_DATA_KEY) {
 326                         fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
 327                                             path->slots[0],
 328                                             struct btrfs_file_extent_item);
 329                         extent_start = btrfs_file_extent_disk_blocknr(fi);
 330                         extent_num_blocks =
 331                                 btrfs_file_extent_disk_num_blocks(fi);
 332                         inode->i_blocks -=
 333                                 btrfs_file_extent_num_blocks(fi) >> 9;
 334                         found_extent = 1;
 335                 } else {
 336                         found_extent = 0;
 337                 }
 338                 ret = btrfs_del_item(trans, root, path);
 339                 BUG_ON(ret);
 340                 btrfs_release_path(root, path);
 341                 if (found_extent) {
 342                         ret = btrfs_free_extent(trans, root, extent_start,
 343                                                 extent_num_blocks, 0);
 344                         BUG_ON(ret);
 345                 }
 346         }
 347         ret = 0;
 348 error:
 349         btrfs_release_path(root, path);
 350         btrfs_free_path(path);
 351         return ret;
 352 }
 353
 354 static void btrfs_delete_inode(struct inode *inode)
 355 {
 356         struct btrfs_trans_handle *trans;
 357         struct btrfs_root *root = BTRFS_I(inode)->root;
 358         int ret;
 359
 360         truncate_inode_pages(&inode->i_data, 0);
 361         if (is_bad_inode(inode)) {
 362                 goto no_delete;
 363         }
 364         inode->i_size = 0;
 365         mutex_lock(&root->fs_info->fs_mutex);
 366         trans = btrfs_start_transaction(root, 1);
 367         if (S_ISREG(inode->i_mode)) {
 368                 ret = btrfs_truncate_in_trans(trans, root, inode);
 369                 BUG_ON(ret);
 370         }
 371         btrfs_free_inode(trans, root, inode);
 372         btrfs_end_transaction(trans, root);
 373         mutex_unlock(&root->fs_info->fs_mutex);
 374         return;
 375 no_delete:
 376         clear_inode(inode);
 377 }
 378
 379 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
 380                                struct btrfs_key *location)
 381 {
 382         const char *name = dentry->d_name.name;
 383         int namelen = dentry->d_name.len;
 384         struct btrfs_dir_item *di;
 385         struct btrfs_path *path;
 386         struct btrfs_root *root = BTRFS_I(dir)->root;
 387         int ret;
 388
 389         path = btrfs_alloc_path();
 390         BUG_ON(!path);
 391         btrfs_init_path(path);
 392         ret = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
 393                                     namelen, 0);
 394         if (ret || !btrfs_match_dir_item_name(root, path, name, namelen)) {
 395                 location->objectid = 0;
 396                 ret = 0;
 397                 goto out;
 398         }
 399         di = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
 400                             struct btrfs_dir_item);
 401         btrfs_disk_key_to_cpu(location, &di->location);
 402 out:
 403         btrfs_release_path(root, path);
 404         btrfs_free_path(path);
 405         return ret;
 406 }
 407
 408 int fixup_tree_root_location(struct btrfs_root *root,
 409                              struct btrfs_key *location,
 410                              struct btrfs_root **sub_root)
 411 {
 412         struct btrfs_path *path;
 413         struct btrfs_root_item *ri;
 414
 415         if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
 416                 return 0;
 417         if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
 418                 return 0;
 419
 420         path = btrfs_alloc_path();
 421         BUG_ON(!path);
 422         mutex_lock(&root->fs_info->fs_mutex);
 423
 424         *sub_root = btrfs_read_fs_root(root->fs_info, location);
 425         if (IS_ERR(*sub_root))
 426                 return PTR_ERR(*sub_root);
 427
 428         ri = &(*sub_root)->root_item;
 429         location->objectid = btrfs_root_dirid(ri);
 430         location->flags = 0;
 431         btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 432         location->offset = 0;
 433
 434         btrfs_free_path(path);
 435         mutex_unlock(&root->fs_info->fs_mutex);
 436         return 0;
 437 }
 438
 439 int btrfs_init_locked_inode(struct inode *inode, void *p)
 440 {
 441         struct btrfs_iget_args *args = p;
 442         inode->i_ino = args->ino;
 443         BTRFS_I(inode)->root = args->root;
 444         return 0;
 445 }
 446
 447 int btrfs_find_actor(struct inode *inode, void *opaque)
 448 {
 449         struct btrfs_iget_args *args = opaque;
 450         return (args->ino == inode->i_ino &&
 451                 args->root == BTRFS_I(inode)->root);
 452 }
 453
 454 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 455                                 struct btrfs_root *root)
 456 {
 457         struct inode *inode;
 458         struct btrfs_iget_args args;
 459         args.ino = objectid;
 460         args.root = root;
 461
 462         inode = iget5_locked(s, objectid, btrfs_find_actor,
 463                              btrfs_init_locked_inode,
 464                              (void *)&args);
 465         return inode;
 466 }
 467
 468 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 469                                    struct nameidata *nd)
 470 {
 471         struct inode * inode;
 472         struct btrfs_inode *bi = BTRFS_I(dir);
 473         struct btrfs_root *root = bi->root;
 474         struct btrfs_root *sub_root = root;
 475         struct btrfs_key location;
 476         int ret;
 477
 478         if (dentry->d_name.len > BTRFS_NAME_LEN)
 479                 return ERR_PTR(-ENAMETOOLONG);
 480         mutex_lock(&root->fs_info->fs_mutex);
 481         ret = btrfs_inode_by_name(dir, dentry, &location);
 482         mutex_unlock(&root->fs_info->fs_mutex);
 483         if (ret < 0)
 484                 return ERR_PTR(ret);
 485         inode = NULL;
 486         if (location.objectid) {
 487                 ret = fixup_tree_root_location(root, &location, &sub_root);
 488                 if (ret < 0)
 489                         return ERR_PTR(ret);
 490                 if (ret > 0)
 491                         return ERR_PTR(-ENOENT);
 492                 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
 493                                           sub_root);
 494                 if (!inode)
 495                         return ERR_PTR(-EACCES);
 496                 if (inode->i_state & I_NEW) {
 497                         if (sub_root != root) {
 498 printk("adding new root for inode %lu root %p (found %p)\n", inode->i_ino, sub_root, BTRFS_I(inode)->root);
 499                                 igrab(inode);
 500                                 sub_root->inode = inode;
 501                         }
 502                         BTRFS_I(inode)->root = sub_root;
 503                         memcpy(&BTRFS_I(inode)->location, &location,
 504                                sizeof(location));
 505                         btrfs_read_locked_inode(inode);
 506                         unlock_new_inode(inode);
 507                 }
 508         }
 509         return d_splice_alias(inode, dentry);
 510 }
 511
 512 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 513 {
 514         struct inode *inode = filp->f_path.dentry->d_inode;
 515         struct btrfs_root *root = BTRFS_I(inode)->root;
 516         struct btrfs_item *item;
 517         struct btrfs_dir_item *di;
 518         struct btrfs_key key;
 519         struct btrfs_path *path;
 520         int ret;
 521         u32 nritems;
 522         struct btrfs_leaf *leaf;
 523         int slot;
 524         int advance;
 525         unsigned char d_type = DT_UNKNOWN;
 526         int over = 0;
 527         int key_type = BTRFS_DIR_INDEX_KEY;
 528
 529         /* FIXME, use a real flag for deciding about the key type */
 530         if (root->fs_info->tree_root == root)
 531                 key_type = BTRFS_DIR_ITEM_KEY;
 532         mutex_lock(&root->fs_info->fs_mutex);
 533         key.objectid = inode->i_ino;
 534         key.flags = 0;
 535         btrfs_set_key_type(&key, key_type);
 536         key.offset = filp->f_pos;
 537         path = btrfs_alloc_path();
 538         btrfs_init_path(path);
 539         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 540         if (ret < 0)
 541                 goto err;
 542         advance = 0;
 543         while(1) {
 544                 leaf = btrfs_buffer_leaf(path->nodes[0]);
 545                 nritems = btrfs_header_nritems(&leaf->header);
 546                 slot = path->slots[0];
 547                 if (advance || slot >= nritems) {
 548                         if (slot >= nritems -1) {
 549                                 ret = btrfs_next_leaf(root, path);
 550                                 if (ret)
 551                                         break;
 552                                 leaf = btrfs_buffer_leaf(path->nodes[0]);
 553                                 nritems = btrfs_header_nritems(&leaf->header);
 554                                 slot = path->slots[0];
 555                         } else {
 556                                 slot++;
 557                                 path->slots[0]++;
 558                         }
 559                 }
 560                 advance = 1;
 561                 item = leaf->items + slot;
 562                 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
 563                         break;
 564                 if (key_type == BTRFS_DIR_INDEX_KEY &&
 565                     btrfs_disk_key_offset(&item->key) > root->highest_inode)
 566                         break;
 567                 if (btrfs_disk_key_type(&item->key) != key_type)
 568                         continue;
 569                 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
 570                         continue;
 571                 filp->f_pos = btrfs_disk_key_offset(&item->key);
 572                 advance = 1;
 573                 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 574                 over = filldir(dirent, (const char *)(di + 1),
 575                                btrfs_dir_name_len(di),
 576                                btrfs_disk_key_offset(&item->key),
 577                                btrfs_disk_key_objectid(&di->location), d_type);
 578                 if (over)
 579                         goto nopos;
 580         }
 581         filp->f_pos++;
 582 nopos:
 583         ret = 0;
 584 err:
 585         btrfs_release_path(root, path);
 586         btrfs_free_path(path);
 587         mutex_unlock(&root->fs_info->fs_mutex);
 588         return ret;
 589 }
 590
 591 static void btrfs_put_super (struct super_block * sb)
 592 {
 593         struct btrfs_root *root = btrfs_sb(sb);
 594         int ret;
 595
 596         ret = close_ctree(root);
 597         if (ret) {
 598                 printk("close ctree returns %d\n", ret);
 599         }
 600         sb->s_fs_info = NULL;
 601 }
 602
 603 static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 604 {
 605         struct inode * inode;
 606         struct dentry * root_dentry;
 607         struct btrfs_super_block *disk_super;
 608         struct btrfs_root *tree_root;
 609         struct btrfs_inode *bi;
 610
 611         sb->s_maxbytes = MAX_LFS_FILESIZE;
 612         sb->s_magic = BTRFS_SUPER_MAGIC;
 613         sb->s_op = &btrfs_super_ops;
 614         sb->s_time_gran = 1;
 615
 616         tree_root = open_ctree(sb);
 617
 618         if (!tree_root) {
 619                 printk("btrfs: open_ctree failed\n");
 620                 return -EIO;
 621         }
 622         sb->s_fs_info = tree_root;
 623         disk_super = tree_root->fs_info->disk_super;
 624         printk("read in super total blocks %Lu root %Lu\n",
 625                btrfs_super_total_blocks(disk_super),
 626                btrfs_super_root_dir(disk_super));
 627
 628         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 629                                   tree_root);
 630         bi = BTRFS_I(inode);
 631         bi->location.objectid = inode->i_ino;
 632         bi->location.offset = 0;
 633         bi->location.flags = 0;
 634         bi->root = tree_root;
 635         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 636
 637         if (!inode)
 638                 return -ENOMEM;
 639         if (inode->i_state & I_NEW) {
 640                 btrfs_read_locked_inode(inode);
 641                 unlock_new_inode(inode);
 642         }
 643
 644         root_dentry = d_alloc_root(inode);
 645         if (!root_dentry) {
 646                 iput(inode);
 647                 return -ENOMEM;
 648         }
 649         sb->s_root = root_dentry;
 650
 651         return 0;
 652 }
 653
 654 static void fill_inode_item(struct btrfs_inode_item *item,
 655                             struct inode *inode)
 656 {
 657         btrfs_set_inode_uid(item, inode->i_uid);
 658         btrfs_set_inode_gid(item, inode->i_gid);
 659         btrfs_set_inode_size(item, inode->i_size);
 660         btrfs_set_inode_mode(item, inode->i_mode);
 661         btrfs_set_inode_nlink(item, inode->i_nlink);
 662         btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
 663         btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
 664         btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
 665         btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
 666         btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
 667         btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
 668         btrfs_set_inode_nblocks(item, inode->i_blocks);
 669         btrfs_set_inode_generation(item, inode->i_generation);
 670 }
 671
 672 static int btrfs_update_inode(struct btrfs_trans_handle *trans,
 673                               struct btrfs_root *root,
 674                               struct inode *inode)
 675 {
 676         struct btrfs_inode_item *inode_item;
 677         struct btrfs_path *path;
 678         int ret;
 679
 680         path = btrfs_alloc_path();
 681         BUG_ON(!path);
 682         btrfs_init_path(path);
 683         ret = btrfs_lookup_inode(trans, root, path,
 684                                  &BTRFS_I(inode)->location, 1);
 685         if (ret) {
 686                 if (ret > 0)
 687                         ret = -ENOENT;
 688                 goto failed;
 689         }
 690
 691         inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
 692                                   path->slots[0],
 693                                   struct btrfs_inode_item);
 694
 695         fill_inode_item(inode_item, inode);
 696         btrfs_mark_buffer_dirty(path->nodes[0]);
 697         ret = 0;
 698 failed:
 699         btrfs_release_path(root, path);
 700         btrfs_free_path(path);
 701         return ret;
 702 }
 703
 704 static int btrfs_write_inode(struct inode *inode, int wait)
 705 {
 706         struct btrfs_root *root = BTRFS_I(inode)->root;
 707         struct btrfs_trans_handle *trans;
 708         int ret;
 709
 710         mutex_lock(&root->fs_info->fs_mutex);
 711         trans = btrfs_start_transaction(root, 1);
 712         ret = btrfs_update_inode(trans, root, inode);
 713         if (wait)
 714                 btrfs_commit_transaction(trans, root);
 715         else
 716                 btrfs_end_transaction(trans, root);
 717         mutex_unlock(&root->fs_info->fs_mutex);
 718         return ret;
 719 }
 720
 721 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 722                                      struct btrfs_root *root,
 723                                      u64 objectid, int mode)
 724 {
 725         struct inode *inode;
 726         struct btrfs_inode_item inode_item;
 727         struct btrfs_key *location;
 728         int ret;
 729
 730         inode = new_inode(root->fs_info->sb);
 731         if (!inode)
 732                 return ERR_PTR(-ENOMEM);
 733
 734         BTRFS_I(inode)->root = root;
 735
 736         inode->i_uid = current->fsuid;
 737         inode->i_gid = current->fsgid;
 738         inode->i_mode = mode;
 739         inode->i_ino = objectid;
 740         inode->i_blocks = 0;
 741         inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 742         fill_inode_item(&inode_item, inode);
 743         location = &BTRFS_I(inode)->location;
 744         location->objectid = objectid;
 745         location->flags = 0;
 746         location->offset = 0;
 747         btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 748
 749         ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
 750         BUG_ON(ret);
 751
 752         insert_inode_hash(inode);
 753         return inode;
 754 }
 755
 756 static int btrfs_add_link(struct btrfs_trans_handle *trans,
 757                             struct dentry *dentry, struct inode *inode)
 758 {
 759         int ret;
 760         struct btrfs_key key;
 761         struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
 762         key.objectid = inode->i_ino;
 763         key.flags = 0;
 764         btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 765         key.offset = 0;
 766
 767         ret = btrfs_insert_dir_item(trans, root,
 768                                     dentry->d_name.name, dentry->d_name.len,
 769                                     dentry->d_parent->d_inode->i_ino,
 770                                     &key, 0);
 771         if (ret == 0) {
 772                 dentry->d_parent->d_inode->i_size += dentry->d_name.len * 2;
 773                 ret = btrfs_update_inode(trans, root,
 774                                          dentry->d_parent->d_inode);
 775         }
 776         return ret;
 777 }
 778
 779 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
 780                             struct dentry *dentry, struct inode *inode)
 781 {
 782         int err = btrfs_add_link(trans, dentry, inode);
 783         if (!err) {
 784                 d_instantiate(dentry, inode);
 785                 return 0;
 786         }
 787         if (err > 0)
 788                 err = -EEXIST;
 789         return err;
 790 }
 791
 792 static int btrfs_create(struct inode *dir, struct dentry *dentry,
 793                         int mode, struct nameidata *nd)
 794 {
 795         struct btrfs_trans_handle *trans;
 796         struct btrfs_root *root = BTRFS_I(dir)->root;
 797         struct inode *inode;
 798         int err;
 799         int drop_inode = 0;
 800         u64 objectid;
 801
 802         mutex_lock(&root->fs_info->fs_mutex);
 803         trans = btrfs_start_transaction(root, 1);
 804
 805         err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
 806         if (err) {
 807                 err = -ENOSPC;
 808                 goto out_unlock;
 809         }
 810
 811         inode = btrfs_new_inode(trans, root, objectid, mode);
 812         err = PTR_ERR(inode);
 813         if (IS_ERR(inode))
 814                 goto out_unlock;
 815         // FIXME mark the inode dirty
 816         err = btrfs_add_nondir(trans, dentry, inode);
 817         if (err)
 818                 drop_inode = 1;
 819         else {
 820                 inode->i_mapping->a_ops = &btrfs_aops;
 821                 inode->i_fop = &btrfs_file_operations;
 822                 inode->i_op = &btrfs_file_inode_operations;
 823         }
 824         dir->i_sb->s_dirt = 1;
 825 out_unlock:
 826         btrfs_end_transaction(trans, root);
 827         mutex_unlock(&root->fs_info->fs_mutex);
 828
 829         if (drop_inode) {
 830                 inode_dec_link_count(inode);
 831                 iput(inode);
 832         }
 833         return err;
 834 }
 835
 836 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
 837                                 struct btrfs_root *root,
 838                                 u64 objectid, u64 dirid)
 839 {
 840         int ret;
 841         char buf[2];
 842         struct btrfs_key key;
 843
 844         buf[0] = '.';
 845         buf[1] = '.';
 846
 847         key.objectid = objectid;
 848         key.offset = 0;
 849         key.flags = 0;
 850         btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 851
 852         ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
 853                                     &key, 1);
 854         if (ret)
 855                 goto error;
 856         key.objectid = dirid;
 857         ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
 858                                     &key, 1);
 859         if (ret)
 860                 goto error;
 861 error:
 862         return ret;
 863 }
 864
 865 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 866 {
 867         struct inode *inode;
 868         struct btrfs_trans_handle *trans;
 869         struct btrfs_root *root = BTRFS_I(dir)->root;
 870         int err = 0;
 871         int drop_on_err = 0;
 872         u64 objectid;
 873
 874         mutex_lock(&root->fs_info->fs_mutex);
 875         trans = btrfs_start_transaction(root, 1);
 876         if (IS_ERR(trans)) {
 877                 err = PTR_ERR(trans);
 878                 goto out_unlock;
 879         }
 880
 881         err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
 882         if (err) {
 883                 err = -ENOSPC;
 884                 goto out_unlock;
 885         }
 886
 887         inode = btrfs_new_inode(trans, root, objectid, S_IFDIR | mode);
 888         if (IS_ERR(inode)) {
 889                 err = PTR_ERR(inode);
 890                 goto out_fail;
 891         }
 892         drop_on_err = 1;
 893         inode->i_op = &btrfs_dir_inode_operations;
 894         inode->i_fop = &btrfs_dir_file_operations;
 895
 896         err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
 897         if (err)
 898                 goto out_fail;
 899
 900         inode->i_size = 6;
 901         err = btrfs_update_inode(trans, root, inode);
 902         if (err)
 903                 goto out_fail;
 904         err = btrfs_add_link(trans, dentry, inode);
 905         if (err)
 906                 goto out_fail;
 907         d_instantiate(dentry, inode);
 908         drop_on_err = 0;
 909
 910 out_fail:
 911         btrfs_end_transaction(trans, root);
 912 out_unlock:
 913         mutex_unlock(&root->fs_info->fs_mutex);
 914         if (drop_on_err)
 915                 iput(inode);
 916         return err;
 917 }
 918
 919 static int btrfs_sync_fs(struct super_block *sb, int wait)
 920 {
 921         struct btrfs_trans_handle *trans;
 922         struct btrfs_root *root;
 923         int ret;
 924         root = btrfs_sb(sb);
 925
 926         sb->s_dirt = 0;
 927         if (!wait) {
 928                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 929                 return 0;
 930         }
 931         filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
 932         mutex_lock(&root->fs_info->fs_mutex);
 933         trans = btrfs_start_transaction(root, 1);
 934         ret = btrfs_commit_transaction(trans, root);
 935         sb->s_dirt = 0;
 936         BUG_ON(ret);
 937 printk("btrfs sync_fs\n");
 938         mutex_unlock(&root->fs_info->fs_mutex);
 939         return 0;
 940 }
 941
 942 #if 0
 943 static int btrfs_get_block_inline(struct inode *inode, sector_t iblock,
 944                            struct buffer_head *result, int create)
 945 {
 946         struct btrfs_root *root = btrfs_sb(inode->i_sb);
 947         struct btrfs_path *path;
 948         struct btrfs_key key;
 949         struct btrfs_leaf *leaf;
 950         int num_bytes = result->b_size;
 951         int item_size;
 952         int ret;
 953         u64 pos;
 954         char *ptr;
 955         int copy_size;
 956         int err = 0;
 957         char *safe_ptr;
 958         char *data_ptr;
 959
 960         path = btrfs_alloc_path();
 961         BUG_ON(!path);
 962
 963         WARN_ON(create);
 964         if (create) {
 965                 return 0;
 966         }
 967         pos = iblock << inode->i_blkbits;
 968         key.objectid = inode->i_ino;
 969         key.flags = 0;
 970         btrfs_set_key_type(&key, BTRFS_INLINE_DATA_KEY);
 971         ptr = kmap(result->b_page);
 972         safe_ptr = ptr;
 973         ptr += (pos & (PAGE_CACHE_SIZE -1));
 974 again:
 975         key.offset = pos;
 976         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 977         if (ret) {
 978                 if (ret < 0)
 979                         err = ret;
 980                 else
 981                         err = 0;
 982                 goto out;
 983         }
 984         leaf = btrfs_buffer_leaf(path->nodes[0]);
 985         item_size = btrfs_item_size(leaf->items + path->slots[0]);
 986         copy_size = min(num_bytes, item_size);
 987         data_ptr = btrfs_item_ptr(leaf, path->slots[0], char);
 988         WARN_ON(safe_ptr + PAGE_CACHE_SIZE < ptr + copy_size);
 989         memcpy(ptr, data_ptr, copy_size);
 990         pos += copy_size;
 991         num_bytes -= copy_size;
 992         WARN_ON(num_bytes < 0);
 993         ptr += copy_size;
 994         btrfs_release_path(root, path);
 995         if (num_bytes != 0) {
 996                 if (pos >= i_size_read(inode))
 997                         memset(ptr, 0, num_bytes);
 998                 else
 999                         goto again;
1000         }
1001         set_buffer_uptodate(result);
1002         map_bh(result, inode->i_sb, 0);
1003         err = 0;
1004 out:
1005         btrfs_free_path(path);
1006         kunmap(result->b_page);
1007         return err;
1008 }
1009 #endif
1010
1011 static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
1012                            struct buffer_head *result, int create)
1013 {
1014         int ret;
1015         int err = 0;
1016         u64 blocknr;
1017         u64 extent_start = 0;
1018         u64 extent_end = 0;
1019         u64 objectid = inode->i_ino;
1020         struct btrfs_path *path;
1021         struct btrfs_root *root = BTRFS_I(inode)->root;
1022         struct btrfs_trans_handle *trans = NULL;
1023         struct btrfs_file_extent_item *item;
1024         struct btrfs_leaf *leaf;
1025         struct btrfs_disk_key *found_key;
1026
1027         path = btrfs_alloc_path();
1028         BUG_ON(!path);
1029         btrfs_init_path(path);
1030         if (create)
1031                 trans = btrfs_start_transaction(root, 1);
1032
1033         ret = btrfs_lookup_file_extent(trans, root, path,
1034                                        inode->i_ino,
1035                                        iblock << inode->i_blkbits, create);
1036         if (ret < 0) {
1037                 err = ret;
1038                 goto out;
1039         }
1040
1041         if (ret != 0) {
1042                 if (path->slots[0] == 0) {
1043                         btrfs_release_path(root, path);
1044                         goto allocate;
1045                 }
1046                 path->slots[0]--;
1047         }
1048
1049         item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1050                               struct btrfs_file_extent_item);
1051         leaf = btrfs_buffer_leaf(path->nodes[0]);
1052         blocknr = btrfs_file_extent_disk_blocknr(item);
1053         blocknr += btrfs_file_extent_offset(item);
1054
1055         /* exact match found, use it, FIXME, deal with extents
1056          * other than the page size
1057          */
1058         if (ret == 0) {
1059                 err = 0;
1060                 BUG_ON(btrfs_file_extent_disk_num_blocks(item) != 1);
1061                 if (create &&
1062                     btrfs_file_extent_generation(item) != trans->transid) {
1063                         struct btrfs_key ins;
1064                         ret = btrfs_alloc_extent(trans, root, 1,
1065                                                  blocknr, (u64)-1, &ins);
1066                         BUG_ON(ret);
1067                         btrfs_set_file_extent_disk_blocknr(item, ins.objectid);
1068                         mark_buffer_dirty(path->nodes[0]);
1069                         ret = btrfs_free_extent(trans, root,
1070                                                 blocknr, 1, 0);
1071                         BUG_ON(ret);
1072                         blocknr = ins.objectid;
1073
1074                 }
1075                 map_bh(result, inode->i_sb, blocknr);
1076                 goto out;
1077         }
1078
1079         /* are we inside the extent that was found? */
1080         found_key = &leaf->items[path->slots[0]].key;
1081         if (btrfs_disk_key_objectid(found_key) != objectid ||
1082             btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY) {
1083                 extent_end = 0;
1084                 extent_start = 0;
1085                 btrfs_release_path(root, path);
1086                 goto allocate;
1087         }
1088
1089         extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1090         extent_start = extent_start >> inode->i_blkbits;
1091         extent_start += btrfs_file_extent_offset(item);
1092         extent_end = extent_start + btrfs_file_extent_num_blocks(item);
1093         if (iblock >= extent_start && iblock < extent_end) {
1094                 err = 0;
1095                 map_bh(result, inode->i_sb, blocknr + iblock - extent_start);
1096                 goto out;
1097         }
1098 allocate:
1099         /* ok, create a new extent */
1100         if (!create) {
1101                 err = 0;
1102                 goto out;
1103         }
1104         ret = btrfs_alloc_file_extent(trans, root, objectid,
1105                                       iblock << inode->i_blkbits,
1106                                       1, extent_end, &blocknr);
1107         if (ret) {
1108                 err = ret;
1109                 goto out;
1110         }
1111         inode->i_blocks += inode->i_sb->s_blocksize >> 9;
1112         set_buffer_new(result);
1113         map_bh(result, inode->i_sb, blocknr);
1114
1115 out:
1116         btrfs_release_path(root, path);
1117         btrfs_free_path(path);
1118         if (trans)
1119                 btrfs_end_transaction(trans, root);
1120         return err;
1121 }
1122
1123 static int btrfs_get_block(struct inode *inode, sector_t iblock,
1124                            struct buffer_head *result, int create)
1125 {
1126         int err;
1127         struct btrfs_root *root = BTRFS_I(inode)->root;
1128         mutex_lock(&root->fs_info->fs_mutex);
1129         err = btrfs_get_block_lock(inode, iblock, result, create);
1130         // err = btrfs_get_block_inline(inode, iblock, result, create);
1131         mutex_unlock(&root->fs_info->fs_mutex);
1132         return err;
1133 }
1134
1135 static int btrfs_prepare_write(struct file *file, struct page *page,
1136                                unsigned from, unsigned to)
1137 {
1138         return nobh_prepare_write(page, from, to, btrfs_get_block);
1139 }
1140 static int btrfs_commit_write(struct file *file, struct page *page,
1141                                unsigned from, unsigned to)
1142 {
1143         return nobh_commit_write(file, page, from, to);
1144 }
1145
1146 static void btrfs_write_super(struct super_block *sb)
1147 {
1148         btrfs_sync_fs(sb, 1);
1149 }
1150
1151 static int btrfs_readpage(struct file *file, struct page *page)
1152 {
1153         return mpage_readpage(page, btrfs_get_block);
1154 }
1155
1156 static int btrfs_readpages(struct file *file, struct address_space *mapping,
1157                            struct list_head *pages, unsigned nr_pages)
1158 {
1159         return mpage_readpages(mapping, pages, nr_pages, btrfs_get_block);
1160 }
1161
1162 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1163 {
1164         return nobh_writepage(page, btrfs_get_block, wbc);
1165 }
1166
1167 static void btrfs_truncate(struct inode *inode)
1168 {
1169         struct btrfs_root *root = BTRFS_I(inode)->root;
1170         int ret;
1171         struct btrfs_trans_handle *trans;
1172
1173         if (!S_ISREG(inode->i_mode))
1174                 return;
1175         if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1176                 return;
1177
1178         nobh_truncate_page(inode->i_mapping, inode->i_size);
1179
1180         /* FIXME, add redo link to tree so we don't leak on crash */
1181         mutex_lock(&root->fs_info->fs_mutex);
1182         trans = btrfs_start_transaction(root, 1);
1183         ret = btrfs_truncate_in_trans(trans, root, inode);
1184         BUG_ON(ret);
1185         ret = btrfs_end_transaction(trans, root);
1186         BUG_ON(ret);
1187         mutex_unlock(&root->fs_info->fs_mutex);
1188         mark_inode_dirty(inode);
1189 }
1190
1191 static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
1192                                 struct page **prepared_pages,
1193                                 const char __user * buf)
1194 {
1195         long page_fault = 0;
1196         int i;
1197         int offset = pos & (PAGE_CACHE_SIZE - 1);
1198
1199         for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
1200                 size_t count = min_t(size_t,
1201                                      PAGE_CACHE_SIZE - offset, write_bytes);
1202                 struct page *page = prepared_pages[i];
1203                 fault_in_pages_readable(buf, count);
1204
1205                 /* Copy data from userspace to the current page */
1206                 kmap(page);
1207                 page_fault = __copy_from_user(page_address(page) + offset,
1208                                               buf, count);
1209                 /* Flush processor's dcache for this page */
1210                 flush_dcache_page(page);
1211                 kunmap(page);
1212                 buf += count;
1213                 write_bytes -= count;
1214
1215                 if (page_fault)
1216                         break;
1217         }
1218         return page_fault ? -EFAULT : 0;
1219 }
1220
1221 static void btrfs_drop_pages(struct page **pages, size_t num_pages)
1222 {
1223         size_t i;
1224         for (i = 0; i < num_pages; i++) {
1225                 if (!pages[i])
1226                         break;
1227                 unlock_page(pages[i]);
1228                 mark_page_accessed(pages[i]);
1229                 page_cache_release(pages[i]);
1230         }
1231 }
1232 static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
1233                                    struct btrfs_root *root,
1234                                    struct file *file,
1235                                    struct page **pages,
1236                                    size_t num_pages,
1237                                    loff_t pos,
1238                                    size_t write_bytes)
1239 {
1240         int i;
1241         int offset;
1242         int err = 0;
1243         int ret;
1244         int this_write;
1245         struct inode *inode = file->f_path.dentry->d_inode;
1246
1247         for (i = 0; i < num_pages; i++) {
1248                 offset = pos & (PAGE_CACHE_SIZE -1);
1249                 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1250                 /* FIXME, one block at a time */
1251
1252                 mutex_lock(&root->fs_info->fs_mutex);
1253                 trans = btrfs_start_transaction(root, 1);
1254                 btrfs_csum_file_block(trans, root, inode->i_ino,
1255                                       pages[i]->index << PAGE_CACHE_SHIFT,
1256                                       kmap(pages[i]), PAGE_CACHE_SIZE);
1257                 kunmap(pages[i]);
1258                 SetPageChecked(pages[i]);
1259                 ret = btrfs_end_transaction(trans, root);
1260                 BUG_ON(ret);
1261                 mutex_unlock(&root->fs_info->fs_mutex);
1262
1263                 ret = nobh_commit_write(file, pages[i], offset,
1264                                          offset + this_write);
1265                 pos += this_write;
1266                 if (ret) {
1267                         err = ret;
1268                         goto failed;
1269                 }
1270                 WARN_ON(this_write > write_bytes);
1271                 write_bytes -= this_write;
1272         }
1273 failed:
1274         return err;
1275 }
1276
1277 static int prepare_pages(struct btrfs_trans_handle *trans,
1278                          struct btrfs_root *root,
1279                          struct file *file,
1280                          struct page **pages,
1281                          size_t num_pages,
1282                          loff_t pos,
1283                          unsigned long first_index,
1284                          unsigned long last_index,
1285                          size_t write_bytes)
1286 {
1287         int i;
1288         unsigned long index = pos >> PAGE_CACHE_SHIFT;
1289         struct inode *inode = file->f_path.dentry->d_inode;
1290         int offset;
1291         int err = 0;
1292         int ret;
1293         int this_write;
1294         loff_t isize = i_size_read(inode);
1295
1296         memset(pages, 0, num_pages * sizeof(struct page *));
1297
1298         for (i = 0; i < num_pages; i++) {
1299                 pages[i] = grab_cache_page(inode->i_mapping, index + i);
1300                 if (!pages[i]) {
1301                         err = -ENOMEM;
1302                         goto failed_release;
1303                 }
1304                 offset = pos & (PAGE_CACHE_SIZE -1);
1305                 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1306                 if (!PageUptodate(pages[i]) &&
1307                    (pages[i]->index == first_index ||
1308                     pages[i]->index == last_index) && pos < isize) {
1309                         ret = mpage_readpage(pages[i], btrfs_get_block);
1310                         BUG_ON(ret);
1311                         lock_page(pages[i]);
1312                 }
1313                 ret = nobh_prepare_write(pages[i], offset,
1314                                          offset + this_write,
1315                                          btrfs_get_block);
1316                 pos += this_write;
1317                 if (ret) {
1318                         err = ret;
1319                         goto failed_truncate;
1320                 }
1321                 WARN_ON(this_write > write_bytes);
1322                 write_bytes -= this_write;
1323         }
1324         return 0;
1325
1326 failed_release:
1327         btrfs_drop_pages(pages, num_pages);
1328         return err;
1329
1330 failed_truncate:
1331         btrfs_drop_pages(pages, num_pages);
1332         if (pos > isize)
1333                 vmtruncate(inode, isize);
1334         return err;
1335 }
1336
1337 static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
1338                                 size_t count, loff_t *ppos)
1339 {
1340         loff_t pos;
1341         size_t num_written = 0;
1342         int err = 0;
1343         int ret = 0;
1344         struct inode *inode = file->f_path.dentry->d_inode;
1345         struct btrfs_root *root = BTRFS_I(inode)->root;
1346         struct page *pages[1];
1347         unsigned long first_index;
1348         unsigned long last_index;
1349
1350         if (file->f_flags & O_DIRECT)
1351                 return -EINVAL;
1352         pos = *ppos;
1353
1354         vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
1355         current->backing_dev_info = inode->i_mapping->backing_dev_info;
1356         err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
1357         if (err)
1358                 goto out;
1359         if (count == 0)
1360                 goto out;
1361         err = remove_suid(file->f_path.dentry);
1362         if (err)
1363                 goto out;
1364         file_update_time(file);
1365         mutex_lock(&inode->i_mutex);
1366         first_index = pos >> PAGE_CACHE_SHIFT;
1367         last_index = (pos + count) >> PAGE_CACHE_SHIFT;
1368         while(count > 0) {
1369                 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
1370                 size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
1371                 size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
1372                                         PAGE_CACHE_SHIFT;
1373                 ret = prepare_pages(NULL, root, file, pages, num_pages,
1374                                     pos, first_index, last_index, write_bytes);
1375                 BUG_ON(ret);
1376                 ret = btrfs_copy_from_user(pos, num_pages,
1377                                            write_bytes, pages, buf);
1378                 BUG_ON(ret);
1379
1380                 ret = dirty_and_release_pages(NULL, root, file, pages,
1381                                               num_pages, pos, write_bytes);
1382                 BUG_ON(ret);
1383                 btrfs_drop_pages(pages, num_pages);
1384
1385                 buf += write_bytes;
1386                 count -= write_bytes;
1387                 pos += write_bytes;
1388                 num_written += write_bytes;
1389
1390                 balance_dirty_pages_ratelimited(inode->i_mapping);
1391                 cond_resched();
1392         }
1393         mutex_unlock(&inode->i_mutex);
1394 out:
1395         *ppos = pos;
1396         current->backing_dev_info = NULL;
1397         return num_written ? num_written : err;
1398 }
1399
1400 #if 0
1401 static ssize_t inline_one_page(struct btrfs_root *root, struct inode *inode,
1402                            struct page *page, loff_t pos,
1403                            size_t offset, size_t write_bytes)
1404 {
1405         struct btrfs_path *path;
1406         struct btrfs_trans_handle *trans;
1407         struct btrfs_key key;
1408         struct btrfs_leaf *leaf;
1409         struct btrfs_key found_key;
1410         int ret;
1411         size_t copy_size = 0;
1412         char *dst = NULL;
1413         int err = 0;
1414         size_t num_written = 0;
1415
1416         path = btrfs_alloc_path();
1417         BUG_ON(!path);
1418         mutex_lock(&root->fs_info->fs_mutex);
1419         trans = btrfs_start_transaction(root, 1);
1420         key.objectid = inode->i_ino;
1421         key.flags = 0;
1422         btrfs_set_key_type(&key, BTRFS_INLINE_DATA_KEY);
1423
1424 again:
1425         key.offset = pos;
1426         ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1427         if (ret < 0) {
1428                 err = ret;
1429                 goto out;
1430         }
1431         if (ret == 0) {
1432                 leaf = btrfs_buffer_leaf(path->nodes[0]);
1433                 btrfs_disk_key_to_cpu(&found_key,
1434                                       &leaf->items[path->slots[0]].key);
1435                 copy_size = btrfs_item_size(leaf->items + path->slots[0]);
1436                 dst = btrfs_item_ptr(leaf, path->slots[0], char);
1437                 copy_size = min(write_bytes, copy_size);
1438                 goto copyit;
1439         } else {
1440                 int slot = path->slots[0];
1441                 if (slot > 0) {
1442                         slot--;
1443                 }
1444                 // FIXME find max key
1445                 leaf = btrfs_buffer_leaf(path->nodes[0]);
1446                 btrfs_disk_key_to_cpu(&found_key,
1447                                       &leaf->items[slot].key);
1448                 if (found_key.objectid != inode->i_ino)
1449                         goto insert;
1450                 if (btrfs_key_type(&found_key) != BTRFS_INLINE_DATA_KEY)
1451                         goto insert;
1452                 copy_size = btrfs_item_size(leaf->items + slot);
1453                 if (found_key.offset + copy_size <= pos)
1454                         goto insert;
1455                 dst = btrfs_item_ptr(leaf, path->slots[0], char);
1456                 dst += pos - found_key.offset;
1457                 copy_size = copy_size - (pos - found_key.offset);
1458                 BUG_ON(copy_size < 0);
1459                 copy_size = min(write_bytes, copy_size);
1460                 WARN_ON(copy_size == 0);
1461                 goto copyit;
1462         }
1463 insert:
1464         btrfs_release_path(root, path);
1465         copy_size = min(write_bytes,
1466                         (size_t)BTRFS_LEAF_DATA_SIZE(root) -
1467                         sizeof(struct btrfs_item) * 4);
1468         ret = btrfs_insert_empty_item(trans, root, path, &key, copy_size);
1469         BUG_ON(ret);
1470         dst = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
1471                              path->slots[0], char);
1472 copyit:
1473         WARN_ON(copy_size == 0);
1474         WARN_ON(dst + copy_size >
1475                 btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
1476                                                  path->slots[0], char) +
1477                 btrfs_item_size(btrfs_buffer_leaf(path->nodes[0])->items +
1478                                                   path->slots[0]));
1479         btrfs_memcpy(root, path->nodes[0]->b_data, dst,
1480                      page_address(page) + offset, copy_size);
1481         mark_buffer_dirty(path->nodes[0]);
1482         btrfs_release_path(root, path);
1483         pos += copy_size;
1484         offset += copy_size;
1485         num_written += copy_size;
1486         write_bytes -= copy_size;
1487         if (write_bytes)
1488                 goto again;
1489 out:
1490         btrfs_free_path(path);
1491         ret = btrfs_end_transaction(trans, root);
1492         BUG_ON(ret);
1493         mutex_unlock(&root->fs_info->fs_mutex);
1494         return num_written ? num_written : err;
1495 }
1496
1497 static ssize_t btrfs_file_inline_write(struct file *file,
1498                                        const char __user *buf,
1499                                        size_t count, loff_t *ppos)
1500 {
1501         loff_t pos;
1502         size_t num_written = 0;
1503         int err = 0;
1504         int ret = 0;
1505         struct inode *inode = file->f_path.dentry->d_inode;
1506         struct btrfs_root *root = BTRFS_I(inode)->root;
1507         unsigned long page_index;
1508
1509         if (file->f_flags & O_DIRECT)
1510                 return -EINVAL;
1511         pos = *ppos;
1512
1513         vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
1514         current->backing_dev_info = inode->i_mapping->backing_dev_info;
1515         err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
1516         if (err)
1517                 goto out;
1518         if (count == 0)
1519                 goto out;
1520         err = remove_suid(file->f_path.dentry);
1521         if (err)
1522                 goto out;
1523         file_update_time(file);
1524         mutex_lock(&inode->i_mutex);
1525         while(count > 0) {
1526                 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
1527                 size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
1528                 struct page *page;
1529
1530                 page_index = pos >> PAGE_CACHE_SHIFT;
1531                 page = grab_cache_page(inode->i_mapping, page_index);
1532                 if (!PageUptodate(page)) {
1533                         ret = mpage_readpage(page, btrfs_get_block);
1534                         BUG_ON(ret);
1535                         lock_page(page);
1536                 }
1537                 ret = btrfs_copy_from_user(pos, 1,
1538                                            write_bytes, &page, buf);
1539                 BUG_ON(ret);
1540                 write_bytes = inline_one_page(root, inode, page, pos,
1541                                       offset, write_bytes);
1542                 SetPageUptodate(page);
1543                 if (write_bytes > 0 && pos + write_bytes > inode->i_size) {
1544                         i_size_write(inode, pos + write_bytes);
1545                         mark_inode_dirty(inode);
1546                 }
1547                 page_cache_release(page);
1548                 unlock_page(page);
1549                 if (write_bytes < 0)
1550                         goto out_unlock;
1551                 buf += write_bytes;
1552                 count -= write_bytes;
1553                 pos += write_bytes;
1554                 num_written += write_bytes;
1555
1556                 balance_dirty_pages_ratelimited(inode->i_mapping);
1557                 cond_resched();
1558         }
1559 out_unlock:
1560         mutex_unlock(&inode->i_mutex);
1561 out:
1562         *ppos = pos;
1563         current->backing_dev_info = NULL;
1564         return num_written ? num_written : err;
1565 }
1566 #endif
1567
1568 static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
1569                         unsigned long offset, unsigned long size)
1570 {
1571         char *kaddr;
1572         unsigned long left, count = desc->count;
1573         struct inode *inode = page->mapping->host;
1574
1575         if (size > count)
1576                 size = count;
1577
1578         if (!PageChecked(page)) {
1579                 /* FIXME, do it per block */
1580                 struct btrfs_root *root = BTRFS_I(inode)->root;
1581                 int ret = btrfs_csum_verify_file_block(root,
1582                                           page->mapping->host->i_ino,
1583                                           page->index << PAGE_CACHE_SHIFT,
1584                                           kmap(page), PAGE_CACHE_SIZE);
1585                 if (ret) {
1586                         printk("failed to verify ino %lu page %lu\n",
1587                                page->mapping->host->i_ino,
1588                                page->index);
1589                         memset(page_address(page), 0, PAGE_CACHE_SIZE);
1590                 }
1591                 SetPageChecked(page);
1592                 kunmap(page);
1593         }
1594         /*
1595          * Faults on the destination of a read are common, so do it before
1596          * taking the kmap.
1597          */
1598         if (!fault_in_pages_writeable(desc->arg.buf, size)) {
1599                 kaddr = kmap_atomic(page, KM_USER0);
1600                 left = __copy_to_user_inatomic(desc->arg.buf,
1601                                                 kaddr + offset, size);
1602                 kunmap_atomic(kaddr, KM_USER0);
1603                 if (left == 0)
1604                         goto success;
1605         }
1606
1607         /* Do it the slow way */
1608         kaddr = kmap(page);
1609         left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
1610         kunmap(page);
1611
1612         if (left) {
1613                 size -= left;
1614                 desc->error = -EFAULT;
1615         }
1616 success:
1617         desc->count = count - size;
1618         desc->written += size;
1619         desc->arg.buf += size;
1620         return size;
1621 }
1622
1623 /**
1624  * btrfs_file_aio_read - filesystem read routine
1625  * @iocb:       kernel I/O control block
1626  * @iov:        io vector request
1627  * @nr_segs:    number of segments in the iovec
1628  * @pos:        current file position
1629  */
1630 static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
1631                                    unsigned long nr_segs, loff_t pos)
1632 {
1633         struct file *filp = iocb->ki_filp;
1634         ssize_t retval;
1635         unsigned long seg;
1636         size_t count;
1637         loff_t *ppos = &iocb->ki_pos;
1638
1639         count = 0;
1640         for (seg = 0; seg < nr_segs; seg++) {
1641                 const struct iovec *iv = &iov[seg];
1642
1643                 /*
1644                  * If any segment has a negative length, or the cumulative
1645                  * length ever wraps negative then return -EINVAL.
1646                  */
1647                 count += iv->iov_len;
1648                 if (unlikely((ssize_t)(count|iv->iov_len) < 0))
1649                         return -EINVAL;
1650                 if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len))
1651                         continue;
1652                 if (seg == 0)
1653                         return -EFAULT;
1654                 nr_segs = seg;
1655                 count -= iv->iov_len;   /* This segment is no good */
1656                 break;
1657         }
1658         retval = 0;
1659         if (count) {
1660                 for (seg = 0; seg < nr_segs; seg++) {
1661                         read_descriptor_t desc;
1662
1663                         desc.written = 0;
1664                         desc.arg.buf = iov[seg].iov_base;
1665                         desc.count = iov[seg].iov_len;
1666                         if (desc.count == 0)
1667                                 continue;
1668                         desc.error = 0;
1669                         do_generic_file_read(filp, ppos, &desc,
1670                                              btrfs_read_actor);
1671                         retval += desc.written;
1672                         if (desc.error) {
1673                                 retval = retval ?: desc.error;
1674                                 break;
1675                         }
1676                 }
1677         }
1678         return retval;
1679 }
1680
1681 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1682 {
1683         struct btrfs_trans_handle *trans;
1684         struct btrfs_key key;
1685         struct btrfs_root_item root_item;
1686         struct btrfs_inode_item *inode_item;
1687         struct buffer_head *subvol;
1688         struct btrfs_leaf *leaf;
1689         struct btrfs_root *new_root;
1690         struct inode *inode;
1691         int ret;
1692         u64 objectid;
1693         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1694
1695         mutex_lock(&root->fs_info->fs_mutex);
1696         trans = btrfs_start_transaction(root, 1);
1697         BUG_ON(!trans);
1698
1699         subvol = btrfs_alloc_free_block(trans, root);
1700         leaf = btrfs_buffer_leaf(subvol);
1701         btrfs_set_header_nritems(&leaf->header, 0);
1702         btrfs_set_header_level(&leaf->header, 0);
1703         btrfs_set_header_blocknr(&leaf->header, subvol->b_blocknr);
1704         btrfs_set_header_generation(&leaf->header, trans->transid);
1705         memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
1706                sizeof(leaf->header.fsid));
1707
1708         inode_item = &root_item.inode;
1709         memset(inode_item, 0, sizeof(*inode_item));
1710         btrfs_set_inode_generation(inode_item, 1);
1711         btrfs_set_inode_size(inode_item, 3);
1712         btrfs_set_inode_nlink(inode_item, 1);
1713         btrfs_set_inode_nblocks(inode_item, 1);
1714         btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
1715
1716         btrfs_set_root_blocknr(&root_item, subvol->b_blocknr);
1717         btrfs_set_root_refs(&root_item, 1);
1718
1719         mark_buffer_dirty(subvol);
1720         brelse(subvol);
1721         subvol = NULL;
1722
1723         ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1724                                        0, &objectid);
1725         BUG_ON(ret);
1726
1727         btrfs_set_root_dirid(&root_item, new_dirid);
1728
1729         key.objectid = objectid;
1730         key.offset = 1;
1731         key.flags = 0;
1732         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1733         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1734                                 &root_item);
1735         BUG_ON(ret);
1736
1737         /*
1738          * insert the directory item
1739          */
1740         key.offset = (u64)-1;
1741         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1742                                     name, namelen,
1743                                     root->fs_info->sb->s_root->d_inode->i_ino,
1744                                     &key, 0);
1745         BUG_ON(ret);
1746
1747         ret = btrfs_commit_transaction(trans, root);
1748         BUG_ON(ret);
1749
1750         new_root = btrfs_read_fs_root(root->fs_info, &key);
1751         BUG_ON(!new_root);
1752
1753         trans = btrfs_start_transaction(new_root, 1);
1754         BUG_ON(!trans);
1755
1756         inode = btrfs_new_inode(trans, new_root, new_dirid, S_IFDIR | 0700);
1757         inode->i_op = &btrfs_dir_inode_operations;
1758         inode->i_fop = &btrfs_dir_file_operations;
1759
1760         ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1761         BUG_ON(ret);
1762
1763         inode->i_nlink = 1;
1764         inode->i_size = 6;
1765         ret = btrfs_update_inode(trans, new_root, inode);
1766         BUG_ON(ret);
1767
1768         ret = btrfs_commit_transaction(trans, new_root);
1769         BUG_ON(ret);
1770
1771         iput(inode);
1772
1773         mutex_unlock(&root->fs_info->fs_mutex);
1774         return 0;
1775 }
1776
1777 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1778 {
1779         struct btrfs_trans_handle *trans;
1780         struct btrfs_key key;
1781         struct btrfs_root_item new_root_item;
1782         int ret;
1783         u64 objectid;
1784
1785         if (!root->ref_cows)
1786                 return -EINVAL;
1787
1788         mutex_lock(&root->fs_info->fs_mutex);
1789         trans = btrfs_start_transaction(root, 1);
1790         BUG_ON(!trans);
1791
1792         ret = btrfs_update_inode(trans, root, root->inode);
1793         BUG_ON(ret);
1794
1795         ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1796                                        0, &objectid);
1797         BUG_ON(ret);
1798
1799         memcpy(&new_root_item, &root->root_item,
1800                sizeof(new_root_item));
1801
1802         key.objectid = objectid;
1803         key.offset = 1;
1804         key.flags = 0;
1805         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1806         btrfs_set_root_blocknr(&new_root_item, root->node->b_blocknr);
1807
1808         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1809                                 &new_root_item);
1810         BUG_ON(ret);
1811
1812 printk("adding snapshot name %.*s root %Lu %Lu %u\n", namelen, name, key.objectid, key.offset, key.flags);
1813
1814         /*
1815          * insert the directory item
1816          */
1817         key.offset = (u64)-1;
1818         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1819                                     name, namelen,
1820                                     root->fs_info->sb->s_root->d_inode->i_ino,
1821                                     &key, 0);
1822
1823         BUG_ON(ret);
1824
1825         ret = btrfs_inc_root_ref(trans, root);
1826         BUG_ON(ret);
1827
1828         ret = btrfs_commit_transaction(trans, root);
1829         BUG_ON(ret);
1830         mutex_unlock(&root->fs_info->fs_mutex);
1831         return 0;
1832 }
1833
1834 static int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
1835                        cmd, unsigned long arg)
1836 {
1837         struct btrfs_root *root = BTRFS_I(inode)->root;
1838         struct btrfs_ioctl_vol_args vol_args;
1839         int ret;
1840         int namelen;
1841         struct btrfs_path *path;
1842         u64 root_dirid;
1843
1844         switch (cmd) {
1845         case BTRFS_IOC_SNAP_CREATE:
1846                 if (copy_from_user(&vol_args,
1847                                    (struct btrfs_ioctl_vol_args __user *)arg,
1848                                    sizeof(vol_args)))
1849                         return -EFAULT;
1850                 namelen = strlen(vol_args.name);
1851                 if (namelen > BTRFS_VOL_NAME_MAX)
1852                         return -EINVAL;
1853                 path = btrfs_alloc_path();
1854                 if (!path)
1855                         return -ENOMEM;
1856                 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
1857                 mutex_lock(&root->fs_info->fs_mutex);
1858                 ret = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
1859                                     path, root_dirid,
1860                                     vol_args.name, namelen, 0);
1861                 mutex_unlock(&root->fs_info->fs_mutex);
1862                 btrfs_free_path(path);
1863                 if (ret == 0)
1864                         return -EEXIST;
1865
1866                 if (root == root->fs_info->tree_root)
1867                         ret = create_subvol(root, vol_args.name, namelen);
1868                 else
1869                         ret = create_snapshot(root, vol_args.name, namelen);
1870                 WARN_ON(ret);
1871                 break;
1872         default:
1873                 return -ENOTTY;
1874         }
1875         return 0;
1876 }
1877
1878 static struct kmem_cache *btrfs_inode_cachep;
1879 struct kmem_cache *btrfs_trans_handle_cachep;
1880 struct kmem_cache *btrfs_transaction_cachep;
1881 struct kmem_cache *btrfs_bit_radix_cachep;
1882 struct kmem_cache *btrfs_path_cachep;
1883
1884 /*
1885  * Called inside transaction, so use GFP_NOFS
1886  */
1887 static struct inode *btrfs_alloc_inode(struct super_block *sb)
1888 {
1889         struct btrfs_inode *ei;
1890
1891         ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
1892         if (!ei)
1893                 return NULL;
1894         return &ei->vfs_inode;
1895 }
1896
1897 static void btrfs_destroy_inode(struct inode *inode)
1898 {
1899         WARN_ON(!list_empty(&inode->i_dentry));
1900         WARN_ON(inode->i_data.nrpages);
1901
1902         kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
1903 }
1904
1905 static void init_once(void * foo, struct kmem_cache * cachep,
1906                       unsigned long flags)
1907 {
1908         struct btrfs_inode *ei = (struct btrfs_inode *) foo;
1909
1910         if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1911             SLAB_CTOR_CONSTRUCTOR) {
1912                 inode_init_once(&ei->vfs_inode);
1913         }
1914 }
1915
1916 static int init_inodecache(void)
1917 {
1918         btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
1919                                              sizeof(struct btrfs_inode),
1920                                              0, (SLAB_RECLAIM_ACCOUNT|
1921                                                 SLAB_MEM_SPREAD),
1922                                              init_once, NULL);
1923         btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
1924                                              sizeof(struct btrfs_trans_handle),
1925                                              0, (SLAB_RECLAIM_ACCOUNT|
1926                                                 SLAB_MEM_SPREAD),
1927                                              NULL, NULL);
1928         btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
1929                                              sizeof(struct btrfs_transaction),
1930                                              0, (SLAB_RECLAIM_ACCOUNT|
1931                                                 SLAB_MEM_SPREAD),
1932                                              NULL, NULL);
1933         btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
1934                                              sizeof(struct btrfs_transaction),
1935                                              0, (SLAB_RECLAIM_ACCOUNT|
1936                                                 SLAB_MEM_SPREAD),
1937                                              NULL, NULL);
1938         btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix",
1939                                              256,
1940                                              0, (SLAB_RECLAIM_ACCOUNT|
1941                                                 SLAB_MEM_SPREAD |
1942                                                 SLAB_DESTROY_BY_RCU),
1943                                              NULL, NULL);
1944         if (btrfs_inode_cachep == NULL || btrfs_trans_handle_cachep == NULL ||
1945             btrfs_transaction_cachep == NULL || btrfs_bit_radix_cachep == NULL)
1946                 return -ENOMEM;
1947         return 0;
1948 }
1949
1950 static void destroy_inodecache(void)
1951 {
1952         kmem_cache_destroy(btrfs_inode_cachep);
1953         kmem_cache_destroy(btrfs_trans_handle_cachep);
1954         kmem_cache_destroy(btrfs_transaction_cachep);
1955         kmem_cache_destroy(btrfs_bit_radix_cachep);
1956         kmem_cache_destroy(btrfs_path_cachep);
1957 }
1958
1959 static int btrfs_get_sb(struct file_system_type *fs_type,
1960         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1961 {
1962         return get_sb_bdev(fs_type, flags, dev_name, data,
1963                            btrfs_fill_super, mnt);
1964 }
1965
1966 static struct file_system_type btrfs_fs_type = {
1967         .owner          = THIS_MODULE,
1968         .name           = "btrfs",
1969         .get_sb         = btrfs_get_sb,
1970         .kill_sb        = kill_block_super,
1971         .fs_flags       = FS_REQUIRES_DEV,
1972 };
1973
1974 static struct super_operations btrfs_super_ops = {
1975         .statfs         = simple_statfs,
1976         .delete_inode   = btrfs_delete_inode,
1977         .put_super      = btrfs_put_super,
1978         .read_inode     = btrfs_read_locked_inode,
1979         .write_super    = btrfs_write_super,
1980         .sync_fs        = btrfs_sync_fs,
1981         .write_inode    = btrfs_write_inode,
1982         .alloc_inode    = btrfs_alloc_inode,
1983         .destroy_inode  = btrfs_destroy_inode,
1984 };
1985
1986 static struct inode_operations btrfs_dir_inode_operations = {
1987         .lookup         = btrfs_lookup,
1988         .create         = btrfs_create,
1989         .unlink         = btrfs_unlink,
1990         .mkdir          = btrfs_mkdir,
1991         .rmdir          = btrfs_rmdir,
1992 };
1993
1994 static struct inode_operations btrfs_dir_ro_inode_operations = {
1995         .lookup         = btrfs_lookup,
1996 };
1997
1998 static struct file_operations btrfs_dir_file_operations = {
1999         .llseek         = generic_file_llseek,
2000         .read           = generic_read_dir,
2001         .readdir        = btrfs_readdir,
2002         .ioctl          = btrfs_ioctl,
2003 };
2004
2005 static struct address_space_operations btrfs_aops = {
2006         .readpage       = btrfs_readpage,
2007         .readpages      = btrfs_readpages,
2008         .writepage      = btrfs_writepage,
2009         .sync_page      = block_sync_page,
2010         .prepare_write  = btrfs_prepare_write,
2011         .commit_write   = btrfs_commit_write,
2012 };
2013
2014 static struct inode_operations btrfs_file_inode_operations = {
2015         .truncate       = btrfs_truncate,
2016 };
2017
2018 static struct file_operations btrfs_file_operations = {
2019         .llseek         = generic_file_llseek,
2020         .read           = do_sync_read,
2021         .aio_read       = btrfs_file_aio_read,
2022         .write          = btrfs_file_write,
2023         .mmap           = generic_file_mmap,
2024         .open           = generic_file_open,
2025         .ioctl          = btrfs_ioctl,
2026 };
2027
2028 static int __init init_btrfs_fs(void)
2029 {
2030         int err;
2031         printk("btrfs loaded!\n");
2032         err = init_inodecache();
2033         if (err)
2034                 return err;
2035         kset_set_kset_s(&btrfs_subsys, fs_subsys);
2036         err = subsystem_register(&btrfs_subsys);
2037         if (err)
2038                 goto out;
2039         return register_filesystem(&btrfs_fs_type);
2040 out:
2041         destroy_inodecache();
2042         return err;
2043 }
2044
2045 static void __exit exit_btrfs_fs(void)
2046 {
2047         destroy_inodecache();
2048         unregister_filesystem(&btrfs_fs_type);
2049         subsystem_unregister(&btrfs_subsys);
2050         printk("btrfs unloaded\n");
2051 }
2052
2053 module_init(init_btrfs_fs)
2054 module_exit(exit_btrfs_fs)
2055
2056 MODULE_LICENSE("GPL");