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                 WARN_ON(1);
1033         }
1034
1035         ret = btrfs_lookup_file_extent(trans, root, path,
1036                                        inode->i_ino,
1037                                        iblock << inode->i_blkbits, create);
1038         if (ret < 0) {
1039                 err = ret;
1040                 goto out;
1041         }
1042
1043         if (ret != 0) {
1044                 if (path->slots[0] == 0) {
1045                         btrfs_release_path(root, path);
1046                         goto allocate;
1047                 }
1048                 path->slots[0]--;
1049         }
1050
1051         item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1052                               struct btrfs_file_extent_item);
1053         leaf = btrfs_buffer_leaf(path->nodes[0]);
1054         blocknr = btrfs_file_extent_disk_blocknr(item);
1055         blocknr += btrfs_file_extent_offset(item);
1056
1057         /* exact match found, use it, FIXME, deal with extents
1058          * other than the page size
1059          */
1060         if (0 && ret == 0) {
1061                 err = 0;
1062                 if (create &&
1063                     btrfs_file_extent_generation(item) != trans->transid) {
1064                         struct btrfs_key ins;
1065                         ret = btrfs_alloc_extent(trans, root, 1,
1066                                                  blocknr, (u64)-1, &ins);
1067                         BUG_ON(ret);
1068                         btrfs_set_file_extent_disk_blocknr(item, ins.objectid);
1069                         mark_buffer_dirty(path->nodes[0]);
1070                         ret = btrfs_free_extent(trans, root,
1071                                                 blocknr, 1, 0);
1072                         BUG_ON(ret);
1073                         blocknr = ins.objectid;
1074
1075                 }
1076                 btrfs_map_bh_to_logical(root, result, blocknr);
1077                 goto out;
1078         }
1079
1080         /* are we inside the extent that was found? */
1081         found_key = &leaf->items[path->slots[0]].key;
1082         if (btrfs_disk_key_objectid(found_key) != objectid ||
1083             btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY) {
1084                 extent_end = 0;
1085                 extent_start = 0;
1086                 btrfs_release_path(root, path);
1087                 goto allocate;
1088         }
1089
1090         extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1091         extent_start = extent_start >> inode->i_blkbits;
1092         extent_start += btrfs_file_extent_offset(item);
1093         extent_end = extent_start + btrfs_file_extent_num_blocks(item);
1094         if (iblock >= extent_start && iblock < extent_end) {
1095                 err = 0;
1096                 btrfs_map_bh_to_logical(root, result, blocknr + iblock -
1097                                         extent_start);
1098                 goto out;
1099         }
1100 allocate:
1101         /* ok, create a new extent */
1102         if (!create) {
1103                 err = 0;
1104                 goto out;
1105         }
1106         ret = btrfs_alloc_file_extent(trans, root, objectid,
1107                                       iblock << inode->i_blkbits,
1108                                       1, extent_end, &blocknr);
1109         if (ret) {
1110                 err = ret;
1111                 goto out;
1112         }
1113         inode->i_blocks += inode->i_sb->s_blocksize >> 9;
1114         set_buffer_new(result);
1115         map_bh(result, inode->i_sb, blocknr);
1116
1117         btrfs_map_bh_to_logical(root, result, blocknr);
1118 out:
1119         btrfs_release_path(root, path);
1120         btrfs_free_path(path);
1121         if (trans)
1122                 btrfs_end_transaction(trans, root);
1123         return err;
1124 }
1125
1126 static int btrfs_get_block(struct inode *inode, sector_t iblock,
1127                            struct buffer_head *result, int create)
1128 {
1129         int err;
1130         struct btrfs_root *root = BTRFS_I(inode)->root;
1131         mutex_lock(&root->fs_info->fs_mutex);
1132         err = btrfs_get_block_lock(inode, iblock, result, create);
1133         // err = btrfs_get_block_inline(inode, iblock, result, create);
1134         mutex_unlock(&root->fs_info->fs_mutex);
1135         return err;
1136 }
1137
1138 static int btrfs_prepare_write(struct file *file, struct page *page,
1139                                unsigned from, unsigned to)
1140 {
1141         return nobh_prepare_write(page, from, to, btrfs_get_block);
1142 }
1143 static int btrfs_commit_write(struct file *file, struct page *page,
1144                                unsigned from, unsigned to)
1145 {
1146         return nobh_commit_write(file, page, from, to);
1147 }
1148
1149 static void btrfs_write_super(struct super_block *sb)
1150 {
1151         btrfs_sync_fs(sb, 1);
1152 }
1153
1154 static int btrfs_readpage(struct file *file, struct page *page)
1155 {
1156         return mpage_readpage(page, btrfs_get_block);
1157 }
1158
1159 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1160 {
1161         return nobh_writepage(page, btrfs_get_block, wbc);
1162 }
1163
1164 static void btrfs_truncate(struct inode *inode)
1165 {
1166         struct btrfs_root *root = BTRFS_I(inode)->root;
1167         int ret;
1168         struct btrfs_trans_handle *trans;
1169
1170         if (!S_ISREG(inode->i_mode))
1171                 return;
1172         if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1173                 return;
1174
1175         nobh_truncate_page(inode->i_mapping, inode->i_size);
1176
1177         /* FIXME, add redo link to tree so we don't leak on crash */
1178         mutex_lock(&root->fs_info->fs_mutex);
1179         trans = btrfs_start_transaction(root, 1);
1180         ret = btrfs_truncate_in_trans(trans, root, inode);
1181         BUG_ON(ret);
1182         ret = btrfs_end_transaction(trans, root);
1183         BUG_ON(ret);
1184         mutex_unlock(&root->fs_info->fs_mutex);
1185         mark_inode_dirty(inode);
1186 }
1187
1188 static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
1189                                 struct page **prepared_pages,
1190                                 const char __user * buf)
1191 {
1192         long page_fault = 0;
1193         int i;
1194         int offset = pos & (PAGE_CACHE_SIZE - 1);
1195
1196         for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
1197                 size_t count = min_t(size_t,
1198                                      PAGE_CACHE_SIZE - offset, write_bytes);
1199                 struct page *page = prepared_pages[i];
1200                 fault_in_pages_readable(buf, count);
1201
1202                 /* Copy data from userspace to the current page */
1203                 kmap(page);
1204                 page_fault = __copy_from_user(page_address(page) + offset,
1205                                               buf, count);
1206                 /* Flush processor's dcache for this page */
1207                 flush_dcache_page(page);
1208                 kunmap(page);
1209                 buf += count;
1210                 write_bytes -= count;
1211
1212                 if (page_fault)
1213                         break;
1214         }
1215         return page_fault ? -EFAULT : 0;
1216 }
1217
1218 static void btrfs_drop_pages(struct page **pages, size_t num_pages)
1219 {
1220         size_t i;
1221         for (i = 0; i < num_pages; i++) {
1222                 if (!pages[i])
1223                         break;
1224                 unlock_page(pages[i]);
1225                 mark_page_accessed(pages[i]);
1226                 page_cache_release(pages[i]);
1227         }
1228 }
1229 static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
1230                                    struct btrfs_root *root,
1231                                    struct file *file,
1232                                    struct page **pages,
1233                                    size_t num_pages,
1234                                    u64 extent_offset,
1235                                    loff_t pos,
1236                                    size_t write_bytes)
1237 {
1238         int i;
1239         int offset;
1240         int err = 0;
1241         int ret;
1242         int this_write;
1243         struct inode *inode = file->f_path.dentry->d_inode;
1244
1245         for (i = 0; i < num_pages; i++) {
1246                 offset = pos & (PAGE_CACHE_SIZE -1);
1247                 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1248                 /* FIXME, one block at a time */
1249
1250                 mutex_lock(&root->fs_info->fs_mutex);
1251                 trans = btrfs_start_transaction(root, 1);
1252                 btrfs_csum_file_block(trans, root, inode->i_ino,
1253                                       pages[i]->index << PAGE_CACHE_SHIFT,
1254                                       extent_offset,
1255                                       kmap(pages[i]), PAGE_CACHE_SIZE);
1256                 kunmap(pages[i]);
1257                 SetPageChecked(pages[i]);
1258                 ret = btrfs_end_transaction(trans, root);
1259                 BUG_ON(ret);
1260                 mutex_unlock(&root->fs_info->fs_mutex);
1261
1262                 ret = nobh_commit_write(file, pages[i], offset,
1263                                          offset + this_write);
1264                 pos += this_write;
1265                 if (ret) {
1266                         err = ret;
1267                         goto failed;
1268                 }
1269                 WARN_ON(this_write > write_bytes);
1270                 write_bytes -= this_write;
1271         }
1272 failed:
1273         return err;
1274 }
1275
1276 static int prepare_pages(struct btrfs_trans_handle *trans,
1277                          struct btrfs_root *root,
1278                          struct file *file,
1279                          struct page **pages,
1280                          size_t num_pages,
1281                          loff_t pos,
1282                          unsigned long first_index,
1283                          unsigned long last_index,
1284                          size_t write_bytes,
1285                          u64 alloc_extent_start)
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         struct buffer_head *bh;
1295         struct buffer_head *head;
1296         loff_t isize = i_size_read(inode);
1297
1298         memset(pages, 0, num_pages * sizeof(struct page *));
1299
1300         for (i = 0; i < num_pages; i++) {
1301                 pages[i] = grab_cache_page(inode->i_mapping, index + i);
1302                 if (!pages[i]) {
1303                         err = -ENOMEM;
1304                         goto failed_release;
1305                 }
1306                 offset = pos & (PAGE_CACHE_SIZE -1);
1307                 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1308                 if (!PageUptodate(pages[i]) &&
1309                    (pages[i]->index == first_index ||
1310                     pages[i]->index == last_index) && pos < isize) {
1311                         ret = mpage_readpage(pages[i], btrfs_get_block);
1312                         BUG_ON(ret);
1313                         lock_page(pages[i]);
1314                 }
1315                 create_empty_buffers(pages[i], root->fs_info->sb->s_blocksize,
1316                                      (1 << BH_Uptodate));
1317                 head = page_buffers(pages[i]);
1318                 bh = head;
1319                 do {
1320                         err = btrfs_map_bh_to_logical(root, bh,
1321                                                       alloc_extent_start);
1322                         BUG_ON(err);
1323                         if (err)
1324                                 goto failed_truncate;
1325                         bh = bh->b_this_page;
1326                         alloc_extent_start++;
1327                 } while (bh != head);
1328                 pos += this_write;
1329                 WARN_ON(this_write > write_bytes);
1330                 write_bytes -= this_write;
1331         }
1332         return 0;
1333
1334 failed_release:
1335         btrfs_drop_pages(pages, num_pages);
1336         return err;
1337
1338 failed_truncate:
1339         btrfs_drop_pages(pages, num_pages);
1340         if (pos > isize)
1341                 vmtruncate(inode, isize);
1342         return err;
1343 }
1344
1345 static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
1346                                 size_t count, loff_t *ppos)
1347 {
1348         loff_t pos;
1349         size_t num_written = 0;
1350         int err = 0;
1351         int ret = 0;
1352         struct inode *inode = file->f_path.dentry->d_inode;
1353         struct btrfs_root *root = BTRFS_I(inode)->root;
1354         struct page *pages[1];
1355         unsigned long first_index;
1356         unsigned long last_index;
1357         u64 start_pos;
1358         u64 num_blocks;
1359         u64 alloc_extent_start;
1360         u64 orig_extent_start;
1361         struct btrfs_trans_handle *trans;
1362
1363         if (file->f_flags & O_DIRECT)
1364                 return -EINVAL;
1365         pos = *ppos;
1366
1367         start_pos = pos & ~(root->blocksize - 1);
1368         /* FIXME */
1369         if (start_pos != pos)
1370                 return -EINVAL;
1371         num_blocks = (count + pos - start_pos + root->blocksize - 1) >>
1372                         inode->i_blkbits;
1373
1374         vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
1375         current->backing_dev_info = inode->i_mapping->backing_dev_info;
1376         err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
1377         if (err)
1378                 goto out;
1379         if (count == 0)
1380                 goto out;
1381         err = remove_suid(file->f_path.dentry);
1382         if (err)
1383                 goto out;
1384         file_update_time(file);
1385         mutex_lock(&inode->i_mutex);
1386         first_index = pos >> PAGE_CACHE_SHIFT;
1387         last_index = (pos + count) >> PAGE_CACHE_SHIFT;
1388
1389         mutex_lock(&root->fs_info->fs_mutex);
1390         trans = btrfs_start_transaction(root, 1);
1391         if (!trans) {
1392                 err = -ENOMEM;
1393                 goto out_unlock;
1394         }
1395         ret = btrfs_alloc_file_extent(trans, root, inode->i_ino,
1396                                       start_pos, num_blocks, 1,
1397                                       &alloc_extent_start);
1398         BUG_ON(ret);
1399
1400         orig_extent_start = start_pos;
1401         ret = btrfs_end_transaction(trans, root);
1402         BUG_ON(ret);
1403         mutex_unlock(&root->fs_info->fs_mutex);
1404
1405         while(count > 0) {
1406                 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
1407                 size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
1408                 size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
1409                                         PAGE_CACHE_SHIFT;
1410                 ret = prepare_pages(NULL, root, file, pages, num_pages,
1411                                     pos, first_index, last_index,
1412                                     write_bytes, alloc_extent_start);
1413                 BUG_ON(ret);
1414                 /* FIXME blocks != pagesize */
1415                 alloc_extent_start += num_pages;
1416                 ret = btrfs_copy_from_user(pos, num_pages,
1417                                            write_bytes, pages, buf);
1418                 BUG_ON(ret);
1419
1420                 ret = dirty_and_release_pages(NULL, root, file, pages,
1421                                               num_pages, orig_extent_start,
1422                                               pos, write_bytes);
1423                 BUG_ON(ret);
1424                 btrfs_drop_pages(pages, num_pages);
1425
1426                 buf += write_bytes;
1427                 count -= write_bytes;
1428                 pos += write_bytes;
1429                 num_written += write_bytes;
1430
1431                 balance_dirty_pages_ratelimited(inode->i_mapping);
1432                 cond_resched();
1433         }
1434 out_unlock:
1435         mutex_unlock(&inode->i_mutex);
1436 out:
1437         *ppos = pos;
1438         current->backing_dev_info = NULL;
1439         return num_written ? num_written : err;
1440 }
1441
1442 #if 0
1443 static ssize_t inline_one_page(struct btrfs_root *root, struct inode *inode,
1444                            struct page *page, loff_t pos,
1445                            size_t offset, size_t write_bytes)
1446 {
1447         struct btrfs_path *path;
1448         struct btrfs_trans_handle *trans;
1449         struct btrfs_key key;
1450         struct btrfs_leaf *leaf;
1451         struct btrfs_key found_key;
1452         int ret;
1453         size_t copy_size = 0;
1454         char *dst = NULL;
1455         int err = 0;
1456         size_t num_written = 0;
1457
1458         path = btrfs_alloc_path();
1459         BUG_ON(!path);
1460         mutex_lock(&root->fs_info->fs_mutex);
1461         trans = btrfs_start_transaction(root, 1);
1462         key.objectid = inode->i_ino;
1463         key.flags = 0;
1464         btrfs_set_key_type(&key, BTRFS_INLINE_DATA_KEY);
1465
1466 again:
1467         key.offset = pos;
1468         ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1469         if (ret < 0) {
1470                 err = ret;
1471                 goto out;
1472         }
1473         if (ret == 0) {
1474                 leaf = btrfs_buffer_leaf(path->nodes[0]);
1475                 btrfs_disk_key_to_cpu(&found_key,
1476                                       &leaf->items[path->slots[0]].key);
1477                 copy_size = btrfs_item_size(leaf->items + path->slots[0]);
1478                 dst = btrfs_item_ptr(leaf, path->slots[0], char);
1479                 copy_size = min(write_bytes, copy_size);
1480                 goto copyit;
1481         } else {
1482                 int slot = path->slots[0];
1483                 if (slot > 0) {
1484                         slot--;
1485                 }
1486                 // FIXME find max key
1487                 leaf = btrfs_buffer_leaf(path->nodes[0]);
1488                 btrfs_disk_key_to_cpu(&found_key,
1489                                       &leaf->items[slot].key);
1490                 if (found_key.objectid != inode->i_ino)
1491                         goto insert;
1492                 if (btrfs_key_type(&found_key) != BTRFS_INLINE_DATA_KEY)
1493                         goto insert;
1494                 copy_size = btrfs_item_size(leaf->items + slot);
1495                 if (found_key.offset + copy_size <= pos)
1496                         goto insert;
1497                 dst = btrfs_item_ptr(leaf, path->slots[0], char);
1498                 dst += pos - found_key.offset;
1499                 copy_size = copy_size - (pos - found_key.offset);
1500                 BUG_ON(copy_size < 0);
1501                 copy_size = min(write_bytes, copy_size);
1502                 WARN_ON(copy_size == 0);
1503                 goto copyit;
1504         }
1505 insert:
1506         btrfs_release_path(root, path);
1507         copy_size = min(write_bytes,
1508                         (size_t)BTRFS_LEAF_DATA_SIZE(root) -
1509                         sizeof(struct btrfs_item) * 4);
1510         ret = btrfs_insert_empty_item(trans, root, path, &key, copy_size);
1511         BUG_ON(ret);
1512         dst = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
1513                              path->slots[0], char);
1514 copyit:
1515         WARN_ON(copy_size == 0);
1516         WARN_ON(dst + copy_size >
1517                 btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
1518                                                  path->slots[0], char) +
1519                 btrfs_item_size(btrfs_buffer_leaf(path->nodes[0])->items +
1520                                                   path->slots[0]));
1521         btrfs_memcpy(root, path->nodes[0]->b_data, dst,
1522                      page_address(page) + offset, copy_size);
1523         mark_buffer_dirty(path->nodes[0]);
1524         btrfs_release_path(root, path);
1525         pos += copy_size;
1526         offset += copy_size;
1527         num_written += copy_size;
1528         write_bytes -= copy_size;
1529         if (write_bytes)
1530                 goto again;
1531 out:
1532         btrfs_free_path(path);
1533         ret = btrfs_end_transaction(trans, root);
1534         BUG_ON(ret);
1535         mutex_unlock(&root->fs_info->fs_mutex);
1536         return num_written ? num_written : err;
1537 }
1538
1539 static ssize_t btrfs_file_inline_write(struct file *file,
1540                                        const char __user *buf,
1541                                        size_t count, loff_t *ppos)
1542 {
1543         loff_t pos;
1544         size_t num_written = 0;
1545         int err = 0;
1546         int ret = 0;
1547         struct inode *inode = file->f_path.dentry->d_inode;
1548         struct btrfs_root *root = BTRFS_I(inode)->root;
1549         unsigned long page_index;
1550
1551         if (file->f_flags & O_DIRECT)
1552                 return -EINVAL;
1553         pos = *ppos;
1554
1555         vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
1556         current->backing_dev_info = inode->i_mapping->backing_dev_info;
1557         err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
1558         if (err)
1559                 goto out;
1560         if (count == 0)
1561                 goto out;
1562         err = remove_suid(file->f_path.dentry);
1563         if (err)
1564                 goto out;
1565         file_update_time(file);
1566         mutex_lock(&inode->i_mutex);
1567         while(count > 0) {
1568                 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
1569                 size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
1570                 struct page *page;
1571
1572                 page_index = pos >> PAGE_CACHE_SHIFT;
1573                 page = grab_cache_page(inode->i_mapping, page_index);
1574                 if (!PageUptodate(page)) {
1575                         ret = mpage_readpage(page, btrfs_get_block);
1576                         BUG_ON(ret);
1577                         lock_page(page);
1578                 }
1579                 ret = btrfs_copy_from_user(pos, 1,
1580                                            write_bytes, &page, buf);
1581                 BUG_ON(ret);
1582                 write_bytes = inline_one_page(root, inode, page, pos,
1583                                       offset, write_bytes);
1584                 SetPageUptodate(page);
1585                 if (write_bytes > 0 && pos + write_bytes > inode->i_size) {
1586                         i_size_write(inode, pos + write_bytes);
1587                         mark_inode_dirty(inode);
1588                 }
1589                 page_cache_release(page);
1590                 unlock_page(page);
1591                 if (write_bytes < 0)
1592                         goto out_unlock;
1593                 buf += write_bytes;
1594                 count -= write_bytes;
1595                 pos += write_bytes;
1596                 num_written += write_bytes;
1597
1598                 balance_dirty_pages_ratelimited(inode->i_mapping);
1599                 cond_resched();
1600         }
1601 out_unlock:
1602         mutex_unlock(&inode->i_mutex);
1603 out:
1604         *ppos = pos;
1605         current->backing_dev_info = NULL;
1606         return num_written ? num_written : err;
1607 }
1608 #endif
1609
1610 static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
1611                         unsigned long offset, unsigned long size)
1612 {
1613         char *kaddr;
1614         unsigned long left, count = desc->count;
1615         struct inode *inode = page->mapping->host;
1616
1617         if (size > count)
1618                 size = count;
1619
1620         if (!PageChecked(page)) {
1621                 /* FIXME, do it per block */
1622                 struct btrfs_root *root = BTRFS_I(inode)->root;
1623                 int ret = btrfs_csum_verify_file_block(root,
1624                                           page->mapping->host->i_ino,
1625                                           page->index << PAGE_CACHE_SHIFT,
1626                                           kmap(page), PAGE_CACHE_SIZE);
1627                 if (ret) {
1628                         printk("failed to verify ino %lu page %lu\n",
1629                                page->mapping->host->i_ino,
1630                                page->index);
1631                         memset(page_address(page), 0, PAGE_CACHE_SIZE);
1632                 }
1633                 SetPageChecked(page);
1634                 kunmap(page);
1635         }
1636         /*
1637          * Faults on the destination of a read are common, so do it before
1638          * taking the kmap.
1639          */
1640         if (!fault_in_pages_writeable(desc->arg.buf, size)) {
1641                 kaddr = kmap_atomic(page, KM_USER0);
1642                 left = __copy_to_user_inatomic(desc->arg.buf,
1643                                                 kaddr + offset, size);
1644                 kunmap_atomic(kaddr, KM_USER0);
1645                 if (left == 0)
1646                         goto success;
1647         }
1648
1649         /* Do it the slow way */
1650         kaddr = kmap(page);
1651         left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
1652         kunmap(page);
1653
1654         if (left) {
1655                 size -= left;
1656                 desc->error = -EFAULT;
1657         }
1658 success:
1659         desc->count = count - size;
1660         desc->written += size;
1661         desc->arg.buf += size;
1662         return size;
1663 }
1664
1665 /**
1666  * btrfs_file_aio_read - filesystem read routine
1667  * @iocb:       kernel I/O control block
1668  * @iov:        io vector request
1669  * @nr_segs:    number of segments in the iovec
1670  * @pos:        current file position
1671  */
1672 static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
1673                                    unsigned long nr_segs, loff_t pos)
1674 {
1675         struct file *filp = iocb->ki_filp;
1676         ssize_t retval;
1677         unsigned long seg;
1678         size_t count;
1679         loff_t *ppos = &iocb->ki_pos;
1680
1681         count = 0;
1682         for (seg = 0; seg < nr_segs; seg++) {
1683                 const struct iovec *iv = &iov[seg];
1684
1685                 /*
1686                  * If any segment has a negative length, or the cumulative
1687                  * length ever wraps negative then return -EINVAL.
1688                  */
1689                 count += iv->iov_len;
1690                 if (unlikely((ssize_t)(count|iv->iov_len) < 0))
1691                         return -EINVAL;
1692                 if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len))
1693                         continue;
1694                 if (seg == 0)
1695                         return -EFAULT;
1696                 nr_segs = seg;
1697                 count -= iv->iov_len;   /* This segment is no good */
1698                 break;
1699         }
1700         retval = 0;
1701         if (count) {
1702                 for (seg = 0; seg < nr_segs; seg++) {
1703                         read_descriptor_t desc;
1704
1705                         desc.written = 0;
1706                         desc.arg.buf = iov[seg].iov_base;
1707                         desc.count = iov[seg].iov_len;
1708                         if (desc.count == 0)
1709                                 continue;
1710                         desc.error = 0;
1711                         do_generic_file_read(filp, ppos, &desc,
1712                                              btrfs_read_actor);
1713                         retval += desc.written;
1714                         if (desc.error) {
1715                                 retval = retval ?: desc.error;
1716                                 break;
1717                         }
1718                 }
1719         }
1720         return retval;
1721 }
1722
1723 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1724 {
1725         struct btrfs_trans_handle *trans;
1726         struct btrfs_key key;
1727         struct btrfs_root_item root_item;
1728         struct btrfs_inode_item *inode_item;
1729         struct buffer_head *subvol;
1730         struct btrfs_leaf *leaf;
1731         struct btrfs_root *new_root;
1732         struct inode *inode;
1733         int ret;
1734         u64 objectid;
1735         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1736
1737         mutex_lock(&root->fs_info->fs_mutex);
1738         trans = btrfs_start_transaction(root, 1);
1739         BUG_ON(!trans);
1740
1741         subvol = btrfs_alloc_free_block(trans, root);
1742         leaf = btrfs_buffer_leaf(subvol);
1743         btrfs_set_header_nritems(&leaf->header, 0);
1744         btrfs_set_header_level(&leaf->header, 0);
1745         btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
1746         btrfs_set_header_generation(&leaf->header, trans->transid);
1747         memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
1748                sizeof(leaf->header.fsid));
1749
1750         inode_item = &root_item.inode;
1751         memset(inode_item, 0, sizeof(*inode_item));
1752         btrfs_set_inode_generation(inode_item, 1);
1753         btrfs_set_inode_size(inode_item, 3);
1754         btrfs_set_inode_nlink(inode_item, 1);
1755         btrfs_set_inode_nblocks(inode_item, 1);
1756         btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
1757
1758         btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
1759         btrfs_set_root_refs(&root_item, 1);
1760
1761         mark_buffer_dirty(subvol);
1762         brelse(subvol);
1763         subvol = NULL;
1764
1765         ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1766                                        0, &objectid);
1767         BUG_ON(ret);
1768
1769         btrfs_set_root_dirid(&root_item, new_dirid);
1770
1771         key.objectid = objectid;
1772         key.offset = 1;
1773         key.flags = 0;
1774         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1775         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1776                                 &root_item);
1777         BUG_ON(ret);
1778
1779         /*
1780          * insert the directory item
1781          */
1782         key.offset = (u64)-1;
1783         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1784                                     name, namelen,
1785                                     root->fs_info->sb->s_root->d_inode->i_ino,
1786                                     &key, 0);
1787         BUG_ON(ret);
1788
1789         ret = btrfs_commit_transaction(trans, root);
1790         BUG_ON(ret);
1791
1792         new_root = btrfs_read_fs_root(root->fs_info, &key);
1793         BUG_ON(!new_root);
1794
1795         trans = btrfs_start_transaction(new_root, 1);
1796         BUG_ON(!trans);
1797
1798         inode = btrfs_new_inode(trans, new_root, new_dirid, S_IFDIR | 0700);
1799         inode->i_op = &btrfs_dir_inode_operations;
1800         inode->i_fop = &btrfs_dir_file_operations;
1801
1802         ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1803         BUG_ON(ret);
1804
1805         inode->i_nlink = 1;
1806         inode->i_size = 6;
1807         ret = btrfs_update_inode(trans, new_root, inode);
1808         BUG_ON(ret);
1809
1810         ret = btrfs_commit_transaction(trans, new_root);
1811         BUG_ON(ret);
1812
1813         iput(inode);
1814
1815         mutex_unlock(&root->fs_info->fs_mutex);
1816         return 0;
1817 }
1818
1819 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1820 {
1821         struct btrfs_trans_handle *trans;
1822         struct btrfs_key key;
1823         struct btrfs_root_item new_root_item;
1824         int ret;
1825         u64 objectid;
1826
1827         if (!root->ref_cows)
1828                 return -EINVAL;
1829
1830         mutex_lock(&root->fs_info->fs_mutex);
1831         trans = btrfs_start_transaction(root, 1);
1832         BUG_ON(!trans);
1833
1834         ret = btrfs_update_inode(trans, root, root->inode);
1835         BUG_ON(ret);
1836
1837         ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1838                                        0, &objectid);
1839         BUG_ON(ret);
1840
1841         memcpy(&new_root_item, &root->root_item,
1842                sizeof(new_root_item));
1843
1844         key.objectid = objectid;
1845         key.offset = 1;
1846         key.flags = 0;
1847         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1848         btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
1849
1850         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1851                                 &new_root_item);
1852         BUG_ON(ret);
1853
1854         /*
1855          * insert the directory item
1856          */
1857         key.offset = (u64)-1;
1858         ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1859                                     name, namelen,
1860                                     root->fs_info->sb->s_root->d_inode->i_ino,
1861                                     &key, 0);
1862
1863         BUG_ON(ret);
1864
1865         ret = btrfs_inc_root_ref(trans, root);
1866         BUG_ON(ret);
1867
1868         ret = btrfs_commit_transaction(trans, root);
1869         BUG_ON(ret);
1870         mutex_unlock(&root->fs_info->fs_mutex);
1871         return 0;
1872 }
1873
1874 static int add_disk(struct btrfs_root *root, char *name, int namelen)
1875 {
1876         struct block_device *bdev;
1877         struct btrfs_path *path;
1878         struct super_block *sb = root->fs_info->sb;
1879         struct btrfs_root *dev_root = root->fs_info->dev_root;
1880         struct btrfs_trans_handle *trans;
1881         struct btrfs_device_item *dev_item;
1882         struct btrfs_key key;
1883         u16 item_size;
1884         u64 num_blocks;
1885         u64 new_blocks;
1886         u64 device_id;
1887         int ret;
1888
1889 printk("adding disk %s\n", name);
1890         path = btrfs_alloc_path();
1891         if (!path)
1892                 return -ENOMEM;
1893         num_blocks = btrfs_super_total_blocks(root->fs_info->disk_super);
1894         bdev = open_bdev_excl(name, O_RDWR, sb);
1895         if (IS_ERR(bdev)) {
1896                 ret = PTR_ERR(bdev);
1897 printk("open bdev excl failed ret %d\n", ret);
1898                 goto out_nolock;
1899         }
1900         set_blocksize(bdev, sb->s_blocksize);
1901         new_blocks = bdev->bd_inode->i_size >> sb->s_blocksize_bits;
1902         key.objectid = num_blocks;
1903         key.offset = new_blocks;
1904         key.flags = 0;
1905         btrfs_set_key_type(&key, BTRFS_DEV_ITEM_KEY);
1906
1907         mutex_lock(&dev_root->fs_info->fs_mutex);
1908         trans = btrfs_start_transaction(dev_root, 1);
1909         item_size = sizeof(*dev_item) + namelen;
1910 printk("insert empty on %Lu %Lu %u size %d\n", num_blocks, new_blocks, key.flags, item_size);
1911         ret = btrfs_insert_empty_item(trans, dev_root, path, &key, item_size);
1912         if (ret) {
1913 printk("insert failed %d\n", ret);
1914                 close_bdev_excl(bdev);
1915                 if (ret > 0)
1916                         ret = -EEXIST;
1917                 goto out;
1918         }
1919         dev_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
1920                                   path->slots[0], struct btrfs_device_item);
1921         btrfs_set_device_pathlen(dev_item, namelen);
1922         memcpy(dev_item + 1, name, namelen);
1923
1924         device_id = btrfs_super_last_device_id(root->fs_info->disk_super) + 1;
1925         btrfs_set_super_last_device_id(root->fs_info->disk_super, device_id);
1926         btrfs_set_device_id(dev_item, device_id);
1927         mark_buffer_dirty(path->nodes[0]);
1928
1929         ret = btrfs_insert_dev_radix(root, bdev, device_id, num_blocks,
1930                                      new_blocks);
1931
1932         if (!ret) {
1933                 btrfs_set_super_total_blocks(root->fs_info->disk_super,
1934                                              num_blocks + new_blocks);
1935                 i_size_write(root->fs_info->btree_inode,
1936                              (num_blocks + new_blocks) <<
1937                              root->fs_info->btree_inode->i_blkbits);
1938         }
1939
1940 out:
1941         ret = btrfs_commit_transaction(trans, dev_root);
1942         BUG_ON(ret);
1943         mutex_unlock(&root->fs_info->fs_mutex);
1944 out_nolock:
1945         btrfs_free_path(path);
1946
1947         return ret;
1948 }
1949
1950 static int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
1951                        cmd, unsigned long arg)
1952 {
1953         struct btrfs_root *root = BTRFS_I(inode)->root;
1954         struct btrfs_ioctl_vol_args vol_args;
1955         int ret = 0;
1956         int namelen;
1957         struct btrfs_path *path;
1958         u64 root_dirid;
1959
1960         switch (cmd) {
1961         case BTRFS_IOC_SNAP_CREATE:
1962                 if (copy_from_user(&vol_args,
1963                                    (struct btrfs_ioctl_vol_args __user *)arg,
1964                                    sizeof(vol_args)))
1965                         return -EFAULT;
1966                 namelen = strlen(vol_args.name);
1967                 if (namelen > BTRFS_VOL_NAME_MAX)
1968                         return -EINVAL;
1969                 path = btrfs_alloc_path();
1970                 if (!path)
1971                         return -ENOMEM;
1972                 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
1973                 mutex_lock(&root->fs_info->fs_mutex);
1974                 ret = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
1975                                     path, root_dirid,
1976                                     vol_args.name, namelen, 0);
1977                 mutex_unlock(&root->fs_info->fs_mutex);
1978                 btrfs_free_path(path);
1979                 if (ret == 0)
1980                         return -EEXIST;
1981
1982                 if (root == root->fs_info->tree_root)
1983                         ret = create_subvol(root, vol_args.name, namelen);
1984                 else
1985                         ret = create_snapshot(root, vol_args.name, namelen);
1986                 WARN_ON(ret);
1987                 break;
1988         case BTRFS_IOC_ADD_DISK:
1989                 if (copy_from_user(&vol_args,
1990                                    (struct btrfs_ioctl_vol_args __user *)arg,
1991                                    sizeof(vol_args)))
1992                         return -EFAULT;
1993                 namelen = strlen(vol_args.name);
1994                 if (namelen > BTRFS_VOL_NAME_MAX)
1995                         return -EINVAL;
1996                 vol_args.name[namelen] = '\0';
1997                 ret = add_disk(root, vol_args.name, namelen);
1998                 break;
1999         default:
2000                 return -ENOTTY;
2001         }
2002         return ret;
2003 }
2004
2005 static struct kmem_cache *btrfs_inode_cachep;
2006 struct kmem_cache *btrfs_trans_handle_cachep;
2007 struct kmem_cache *btrfs_transaction_cachep;
2008 struct kmem_cache *btrfs_bit_radix_cachep;
2009 struct kmem_cache *btrfs_path_cachep;
2010
2011 /*
2012  * Called inside transaction, so use GFP_NOFS
2013  */
2014 static struct inode *btrfs_alloc_inode(struct super_block *sb)
2015 {
2016         struct btrfs_inode *ei;
2017
2018         ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2019         if (!ei)
2020                 return NULL;
2021         return &ei->vfs_inode;
2022 }
2023
2024 static void btrfs_destroy_inode(struct inode *inode)
2025 {
2026         WARN_ON(!list_empty(&inode->i_dentry));
2027         WARN_ON(inode->i_data.nrpages);
2028
2029         kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2030 }
2031
2032 static void init_once(void * foo, struct kmem_cache * cachep,
2033                       unsigned long flags)
2034 {
2035         struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2036
2037         if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2038             SLAB_CTOR_CONSTRUCTOR) {
2039                 inode_init_once(&ei->vfs_inode);
2040         }
2041 }
2042
2043 static int init_inodecache(void)
2044 {
2045         btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
2046                                              sizeof(struct btrfs_inode),
2047                                              0, (SLAB_RECLAIM_ACCOUNT|
2048                                                 SLAB_MEM_SPREAD),
2049                                              init_once, NULL);
2050         btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
2051                                              sizeof(struct btrfs_trans_handle),
2052                                              0, (SLAB_RECLAIM_ACCOUNT|
2053                                                 SLAB_MEM_SPREAD),
2054                                              NULL, NULL);
2055         btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
2056                                              sizeof(struct btrfs_transaction),
2057                                              0, (SLAB_RECLAIM_ACCOUNT|
2058                                                 SLAB_MEM_SPREAD),
2059                                              NULL, NULL);
2060         btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
2061                                              sizeof(struct btrfs_transaction),
2062                                              0, (SLAB_RECLAIM_ACCOUNT|
2063                                                 SLAB_MEM_SPREAD),
2064                                              NULL, NULL);
2065         btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix",
2066                                              256,
2067                                              0, (SLAB_RECLAIM_ACCOUNT|
2068                                                 SLAB_MEM_SPREAD |
2069                                                 SLAB_DESTROY_BY_RCU),
2070                                              NULL, NULL);
2071         if (btrfs_inode_cachep == NULL || btrfs_trans_handle_cachep == NULL ||
2072             btrfs_transaction_cachep == NULL || btrfs_bit_radix_cachep == NULL)
2073                 return -ENOMEM;
2074         return 0;
2075 }
2076
2077 static void destroy_inodecache(void)
2078 {
2079         kmem_cache_destroy(btrfs_inode_cachep);
2080         kmem_cache_destroy(btrfs_trans_handle_cachep);
2081         kmem_cache_destroy(btrfs_transaction_cachep);
2082         kmem_cache_destroy(btrfs_bit_radix_cachep);
2083         kmem_cache_destroy(btrfs_path_cachep);
2084 }
2085
2086 static int btrfs_get_sb(struct file_system_type *fs_type,
2087         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2088 {
2089         return get_sb_bdev(fs_type, flags, dev_name, data,
2090                            btrfs_fill_super, mnt);
2091 }
2092
2093 static struct file_system_type btrfs_fs_type = {
2094         .owner          = THIS_MODULE,
2095         .name           = "btrfs",
2096         .get_sb         = btrfs_get_sb,
2097         .kill_sb        = kill_block_super,
2098         .fs_flags       = FS_REQUIRES_DEV,
2099 };
2100
2101 static struct super_operations btrfs_super_ops = {
2102         .statfs         = simple_statfs,
2103         .delete_inode   = btrfs_delete_inode,
2104         .put_super      = btrfs_put_super,
2105         .read_inode     = btrfs_read_locked_inode,
2106         .write_super    = btrfs_write_super,
2107         .sync_fs        = btrfs_sync_fs,
2108         .write_inode    = btrfs_write_inode,
2109         .alloc_inode    = btrfs_alloc_inode,
2110         .destroy_inode  = btrfs_destroy_inode,
2111 };
2112
2113 static struct inode_operations btrfs_dir_inode_operations = {
2114         .lookup         = btrfs_lookup,
2115         .create         = btrfs_create,
2116         .unlink         = btrfs_unlink,
2117         .mkdir          = btrfs_mkdir,
2118         .rmdir          = btrfs_rmdir,
2119 };
2120
2121 static struct inode_operations btrfs_dir_ro_inode_operations = {
2122         .lookup         = btrfs_lookup,
2123 };
2124
2125 static struct file_operations btrfs_dir_file_operations = {
2126         .llseek         = generic_file_llseek,
2127         .read           = generic_read_dir,
2128         .readdir        = btrfs_readdir,
2129         .ioctl          = btrfs_ioctl,
2130 };
2131
2132 static struct address_space_operations btrfs_aops = {
2133         .readpage       = btrfs_readpage,
2134         .writepage      = btrfs_writepage,
2135         .sync_page      = block_sync_page,
2136         .prepare_write  = btrfs_prepare_write,
2137         .commit_write   = btrfs_commit_write,
2138 };
2139
2140 static struct inode_operations btrfs_file_inode_operations = {
2141         .truncate       = btrfs_truncate,
2142 };
2143
2144 static struct file_operations btrfs_file_operations = {
2145         .llseek         = generic_file_llseek,
2146         .read           = do_sync_read,
2147         .aio_read       = btrfs_file_aio_read,
2148         .write          = btrfs_file_write,
2149         .mmap           = generic_file_mmap,
2150         .open           = generic_file_open,
2151         .ioctl          = btrfs_ioctl,
2152 };
2153
2154 static int __init init_btrfs_fs(void)
2155 {
2156         int err;
2157         printk("btrfs loaded!\n");
2158         err = init_inodecache();
2159         if (err)
2160                 return err;
2161         kset_set_kset_s(&btrfs_subsys, fs_subsys);
2162         err = subsystem_register(&btrfs_subsys);
2163         if (err)
2164                 goto out;
2165         return register_filesystem(&btrfs_fs_type);
2166 out:
2167         destroy_inodecache();
2168         return err;
2169 }
2170
2171 static void __exit exit_btrfs_fs(void)
2172 {
2173         destroy_inodecache();
2174         unregister_filesystem(&btrfs_fs_type);
2175         subsystem_unregister(&btrfs_subsys);
2176         printk("btrfs unloaded\n");
2177 }
2178
2179 module_init(init_btrfs_fs)
2180 module_exit(exit_btrfs_fs)
2181
2182 MODULE_LICENSE("GPL");