Btrfs: Use balance_dirty_pages_nr on btree blocks
[linux-2.6-block.git] / fs / btrfs / extent-tree.c
index 49f7cd6e067c801b091377532674e372cea4f772..fe772f9b06cf09fd4b6c730307e7fe3478f8697f 100644 (file)
-#include <linux/module.h>
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/sched.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
 #include "transaction.h"
 
-static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-                           *orig_root, u64 num_blocks, u64 search_start, u64
-                           search_end, struct btrfs_key *ins);
 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
                                 btrfs_root *extent_root);
 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
                               btrfs_root *extent_root);
 
+static int cache_block_group(struct btrfs_root *root,
+                            struct btrfs_block_group_cache *block_group)
+{
+       struct btrfs_path *path;
+       int ret;
+       struct btrfs_key key;
+       struct btrfs_leaf *leaf;
+       struct radix_tree_root *extent_radix;
+       int slot;
+       u64 i;
+       u64 last = 0;
+       u64 hole_size;
+       u64 first_free;
+       int found = 0;
+
+       root = root->fs_info->extent_root;
+       extent_radix = &root->fs_info->extent_map_radix;
+
+       if (block_group->cached)
+               return 0;
+       if (block_group->data)
+               return 0;
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       path->reada = 2;
+       first_free = block_group->key.objectid;
+       key.objectid = block_group->key.objectid;
+       key.flags = 0;
+       key.offset = 0;
+
+       btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
+       ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+
+       if (ret < 0)
+               return ret;
+
+       if (ret && path->slots[0] > 0)
+               path->slots[0]--;
+
+       while(1) {
+               leaf = btrfs_buffer_leaf(path->nodes[0]);
+               slot = path->slots[0];
+               if (slot >= btrfs_header_nritems(&leaf->header)) {
+                       ret = btrfs_next_leaf(root, path);
+                       if (ret < 0)
+                               goto err;
+                       if (ret == 0) {
+                               continue;
+                       } else {
+                               break;
+                       }
+               }
+
+               btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+               if (key.objectid < block_group->key.objectid) {
+                       if (key.objectid + key.offset > first_free)
+                               first_free = key.objectid + key.offset;
+                       goto next;
+               }
+
+               if (key.objectid >= block_group->key.objectid +
+                   block_group->key.offset) {
+                       break;
+               }
+
+               if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
+                       if (!found) {
+                               last = first_free;
+                               found = 1;
+                       }
+                       hole_size = key.objectid - last;
+                       for (i = 0; i < hole_size; i++) {
+                               set_radix_bit(extent_radix, last + i);
+                       }
+                       last = key.objectid + key.offset;
+               }
+next:
+               path->slots[0]++;
+       }
+
+       if (!found)
+               last = first_free;
+       if (block_group->key.objectid +
+           block_group->key.offset > last) {
+               hole_size = block_group->key.objectid +
+                       block_group->key.offset - last;
+               for (i = 0; i < hole_size; i++) {
+                       set_radix_bit(extent_radix,
+                                       last + i);
+               }
+       }
+       block_group->cached = 1;
+err:
+       btrfs_free_path(path);
+       return 0;
+}
+
+struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
+                                                        btrfs_fs_info *info,
+                                                        u64 blocknr)
+{
+       struct btrfs_block_group_cache *block_group;
+       int ret;
+
+       ret = radix_tree_gang_lookup(&info->block_group_radix,
+                                    (void **)&block_group,
+                                    blocknr, 1);
+       if (ret) {
+               if (block_group->key.objectid <= blocknr && blocknr <=
+                   block_group->key.objectid + block_group->key.offset)
+                       return block_group;
+       }
+       ret = radix_tree_gang_lookup(&info->block_group_data_radix,
+                                    (void **)&block_group,
+                                    blocknr, 1);
+       if (ret) {
+               if (block_group->key.objectid <= blocknr && blocknr <=
+                   block_group->key.objectid + block_group->key.offset)
+                       return block_group;
+       }
+       return NULL;
+}
+
+static u64 leaf_range(struct btrfs_root *root)
+{
+       u64 size = BTRFS_LEAF_DATA_SIZE(root);
+       do_div(size, sizeof(struct btrfs_extent_item) +
+               sizeof(struct btrfs_item));
+       return size;
+}
+
+static u64 find_search_start(struct btrfs_root *root,
+                            struct btrfs_block_group_cache **cache_ret,
+                            u64 search_start, int num)
+{
+       unsigned long gang[8];
+       int ret;
+       struct btrfs_block_group_cache *cache = *cache_ret;
+       u64 last = max(search_start, cache->key.objectid);
+
+       if (cache->data)
+               goto out;
+again:
+       ret = cache_block_group(root, cache);
+       if (ret)
+               goto out;
+       while(1) {
+               ret = find_first_radix_bit(&root->fs_info->extent_map_radix,
+                                          gang, last, ARRAY_SIZE(gang));
+               if (!ret)
+                       goto out;
+               last = gang[ret-1] + 1;
+               if (num > 1) {
+                       if (ret != ARRAY_SIZE(gang)) {
+                               goto new_group;
+                       }
+                       if (gang[ret-1] - gang[0] > leaf_range(root)) {
+                               continue;
+                       }
+               }
+               if (gang[0] >= cache->key.objectid + cache->key.offset) {
+                       goto new_group;
+               }
+               return gang[0];
+       }
+out:
+       return max(cache->last_alloc, search_start);
+
+new_group:
+       cache = btrfs_lookup_block_group(root->fs_info,
+                                        last + cache->key.offset - 1);
+       if (!cache) {
+               return max((*cache_ret)->last_alloc, search_start);
+       }
+       cache = btrfs_find_block_group(root, cache,
+                                      last + cache->key.offset - 1, 0, 0);
+       *cache_ret = cache;
+       goto again;
+}
+
+static u64 div_factor(u64 num, int factor)
+{
+       num *= factor;
+       do_div(num, 10);
+       return num;
+}
+
+struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
+                                                struct btrfs_block_group_cache
+                                                *hint, u64 search_start,
+                                                int data, int owner)
+{
+       struct btrfs_block_group_cache *cache[8];
+       struct btrfs_block_group_cache *found_group = NULL;
+       struct btrfs_fs_info *info = root->fs_info;
+       struct radix_tree_root *radix;
+       struct radix_tree_root *swap_radix;
+       u64 used;
+       u64 last = 0;
+       u64 hint_last;
+       int i;
+       int ret;
+       int full_search = 0;
+       int factor = 8;
+       int data_swap = 0;
+
+       if (!owner)
+               factor = 5;
+
+       if (data) {
+               radix = &info->block_group_data_radix;
+               swap_radix = &info->block_group_radix;
+       } else {
+               radix = &info->block_group_radix;
+               swap_radix = &info->block_group_data_radix;
+       }
+
+       if (search_start) {
+               struct btrfs_block_group_cache *shint;
+               shint = btrfs_lookup_block_group(info, search_start);
+               if (shint && shint->data == data) {
+                       used = btrfs_block_group_used(&shint->item);
+                       if (used + shint->pinned <
+                           div_factor(shint->key.offset, factor)) {
+                               return shint;
+                       }
+               }
+       }
+       if (hint && hint->data == data) {
+               used = btrfs_block_group_used(&hint->item);
+               if (used + hint->pinned <
+                   div_factor(hint->key.offset, factor)) {
+                       return hint;
+               }
+               if (used >= div_factor(hint->key.offset, 8)) {
+                       radix_tree_tag_clear(radix,
+                                            hint->key.objectid +
+                                            hint->key.offset - 1,
+                                            BTRFS_BLOCK_GROUP_AVAIL);
+               }
+               last = hint->key.offset * 3;
+               if (hint->key.objectid >= last)
+                       last = max(search_start + hint->key.offset - 1,
+                                  hint->key.objectid - last);
+               else
+                       last = hint->key.objectid + hint->key.offset;
+               hint_last = last;
+       } else {
+               if (hint)
+                       hint_last = max(hint->key.objectid, search_start);
+               else
+                       hint_last = search_start;
+
+               last = hint_last;
+       }
+       while(1) {
+               ret = radix_tree_gang_lookup_tag(radix, (void **)cache,
+                                                last, ARRAY_SIZE(cache),
+                                                BTRFS_BLOCK_GROUP_AVAIL);
+               if (!ret)
+                       break;
+               for (i = 0; i < ret; i++) {
+                       last = cache[i]->key.objectid +
+                               cache[i]->key.offset;
+                       used = btrfs_block_group_used(&cache[i]->item);
+                       if (used + cache[i]->pinned <
+                           div_factor(cache[i]->key.offset, factor)) {
+                               found_group = cache[i];
+                               goto found;
+                       }
+                       if (used >= div_factor(cache[i]->key.offset, 8)) {
+                               radix_tree_tag_clear(radix,
+                                                    cache[i]->key.objectid +
+                                                    cache[i]->key.offset - 1,
+                                                    BTRFS_BLOCK_GROUP_AVAIL);
+                       }
+               }
+               cond_resched();
+       }
+       last = hint_last;
+again:
+       while(1) {
+               ret = radix_tree_gang_lookup(radix, (void **)cache,
+                                            last, ARRAY_SIZE(cache));
+               if (!ret)
+                       break;
+               for (i = 0; i < ret; i++) {
+                       last = cache[i]->key.objectid +
+                               cache[i]->key.offset;
+                       used = btrfs_block_group_used(&cache[i]->item);
+                       if (used + cache[i]->pinned < cache[i]->key.offset) {
+                               found_group = cache[i];
+                               goto found;
+                       }
+                       if (used >= cache[i]->key.offset) {
+                               radix_tree_tag_clear(radix,
+                                                    cache[i]->key.objectid +
+                                                    cache[i]->key.offset - 1,
+                                                    BTRFS_BLOCK_GROUP_AVAIL);
+                       }
+               }
+               cond_resched();
+       }
+       if (!full_search) {
+               last = search_start;
+               full_search = 1;
+               goto again;
+       }
+       if (!data_swap) {
+               struct radix_tree_root *tmp = radix;
+               data_swap = 1;
+               radix = swap_radix;
+               swap_radix = tmp;
+               last = search_start;
+               goto again;
+       }
+       if (!found_group) {
+               ret = radix_tree_gang_lookup(radix,
+                                            (void **)&found_group, 0, 1);
+               if (ret == 0) {
+                       ret = radix_tree_gang_lookup(swap_radix,
+                                                    (void **)&found_group,
+                                                    0, 1);
+               }
+               BUG_ON(ret != 1);
+       }
+found:
+       return found_group;
+}
+
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
                                struct btrfs_root *root,
                                u64 blocknr, u64 num_blocks)
@@ -21,22 +368,21 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
        struct btrfs_key key;
        struct btrfs_leaf *l;
        struct btrfs_extent_item *item;
-       struct btrfs_key ins;
        u32 refs;
 
-       find_free_extent(trans, root->fs_info->extent_root, 0, 0, (u64)-1,
-                        &ins);
        path = btrfs_alloc_path();
-       BUG_ON(!path);
-       btrfs_init_path(path);
+       if (!path)
+               return -ENOMEM;
+
        key.objectid = blocknr;
        key.flags = 0;
        btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
        key.offset = num_blocks;
        ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
                                0, 1);
+       if (ret < 0)
+               return ret;
        if (ret != 0) {
-printk("can't find block %Lu %Lu\n", blocknr, num_blocks);
                BUG();
        }
        BUG_ON(ret != 0);
@@ -53,6 +399,14 @@ printk("can't find block %Lu %Lu\n", blocknr, num_blocks);
        return 0;
 }
 
+int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
+                        struct btrfs_root *root)
+{
+       finish_current_insert(trans, root->fs_info->extent_root);
+       del_pending_extents(trans, root->fs_info->extent_root);
+       return 0;
+}
+
 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
                             struct btrfs_root *root, u64 blocknr,
                             u64 num_blocks, u32 *refs)
@@ -64,19 +418,20 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
        struct btrfs_extent_item *item;
 
        path = btrfs_alloc_path();
-       btrfs_init_path(path);
        key.objectid = blocknr;
        key.offset = num_blocks;
        key.flags = 0;
        btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
        ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
                                0, 0);
+       if (ret < 0)
+               goto out;
        if (ret != 0)
                BUG();
        l = btrfs_buffer_leaf(path->nodes[0]);
        item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
        *refs = btrfs_extent_refs(item);
-       btrfs_release_path(root->fs_info->extent_root, path);
+out:
        btrfs_free_path(path);
        return 0;
 }
@@ -98,6 +453,8 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
        int i;
        int leaf;
        int ret;
+       int faili;
+       int err;
 
        if (!root->ref_cows)
                return 0;
@@ -106,6 +463,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
        buf_leaf = btrfs_buffer_leaf(buf);
        for (i = 0; i < btrfs_header_nritems(&buf_node->header); i++) {
                if (leaf) {
+                       u64 disk_blocknr;
                        key = &buf_leaf->items[i].key;
                        if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
                                continue;
@@ -114,30 +472,268 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
                        if (btrfs_file_extent_type(fi) ==
                            BTRFS_FILE_EXTENT_INLINE)
                                continue;
-                       ret = btrfs_inc_extent_ref(trans, root,
-                                   btrfs_file_extent_disk_blocknr(fi),
+                       disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+                       if (disk_blocknr == 0)
+                               continue;
+                       ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
                                    btrfs_file_extent_disk_num_blocks(fi));
-                       BUG_ON(ret);
+                       if (ret) {
+                               faili = i;
+                               goto fail;
+                       }
                } else {
                        blocknr = btrfs_node_blockptr(buf_node, i);
                        ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
-                       BUG_ON(ret);
+                       if (ret) {
+                               faili = i;
+                               goto fail;
+                       }
+               }
+       }
+       return 0;
+fail:
+       WARN_ON(1);
+       for (i =0; i < faili; i++) {
+               if (leaf) {
+                       u64 disk_blocknr;
+                       key = &buf_leaf->items[i].key;
+                       if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+                               continue;
+                       fi = btrfs_item_ptr(buf_leaf, i,
+                                           struct btrfs_file_extent_item);
+                       if (btrfs_file_extent_type(fi) ==
+                           BTRFS_FILE_EXTENT_INLINE)
+                               continue;
+                       disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+                       if (disk_blocknr == 0)
+                               continue;
+                       err = btrfs_free_extent(trans, root, disk_blocknr,
+                                   btrfs_file_extent_disk_num_blocks(fi), 0);
+                       BUG_ON(err);
+               } else {
+                       blocknr = btrfs_node_blockptr(buf_node, i);
+                       err = btrfs_free_extent(trans, root, blocknr, 1, 0);
+                       BUG_ON(err);
+               }
+       }
+       return ret;
+}
+
+static int write_one_cache_group(struct btrfs_trans_handle *trans,
+                                struct btrfs_root *root,
+                                struct btrfs_path *path,
+                                struct btrfs_block_group_cache *cache)
+{
+       int ret;
+       int pending_ret;
+       struct btrfs_root *extent_root = root->fs_info->extent_root;
+       struct btrfs_block_group_item *bi;
+
+       ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
+       if (ret < 0)
+               goto fail;
+       BUG_ON(ret);
+       bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+                           struct btrfs_block_group_item);
+       memcpy(bi, &cache->item, sizeof(*bi));
+       btrfs_mark_buffer_dirty(path->nodes[0]);
+       btrfs_release_path(extent_root, path);
+fail:
+       finish_current_insert(trans, extent_root);
+       pending_ret = del_pending_extents(trans, extent_root);
+       if (ret)
+               return ret;
+       if (pending_ret)
+               return pending_ret;
+       if (cache->data)
+               cache->last_alloc = cache->first_free;
+       return 0;
+
+}
+
+static int write_dirty_block_radix(struct btrfs_trans_handle *trans,
+                                  struct btrfs_root *root,
+                                  struct radix_tree_root *radix)
+{
+       struct btrfs_block_group_cache *cache[8];
+       int ret;
+       int err = 0;
+       int werr = 0;
+       int i;
+       struct btrfs_path *path;
+       unsigned long off = 0;
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       while(1) {
+               ret = radix_tree_gang_lookup_tag(radix, (void **)cache,
+                                                off, ARRAY_SIZE(cache),
+                                                BTRFS_BLOCK_GROUP_DIRTY);
+               if (!ret)
+                       break;
+               for (i = 0; i < ret; i++) {
+                       err = write_one_cache_group(trans, root,
+                                                   path, cache[i]);
+                       /*
+                        * if we fail to write the cache group, we want
+                        * to keep it marked dirty in hopes that a later
+                        * write will work
+                        */
+                       if (err) {
+                               werr = err;
+                               off = cache[i]->key.objectid +
+                                       cache[i]->key.offset;
+                               continue;
+                       }
+
+                       radix_tree_tag_clear(radix, cache[i]->key.objectid +
+                                            cache[i]->key.offset - 1,
+                                            BTRFS_BLOCK_GROUP_DIRTY);
+               }
+       }
+       btrfs_free_path(path);
+       return werr;
+}
+
+int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
+                                  struct btrfs_root *root)
+{
+       int ret;
+       int ret2;
+       ret = write_dirty_block_radix(trans, root,
+                                     &root->fs_info->block_group_radix);
+       ret2 = write_dirty_block_radix(trans, root,
+                                     &root->fs_info->block_group_data_radix);
+       if (ret)
+               return ret;
+       if (ret2)
+               return ret2;
+       return 0;
+}
+
+static int update_block_group(struct btrfs_trans_handle *trans,
+                             struct btrfs_root *root,
+                             u64 blocknr, u64 num, int alloc, int mark_free,
+                             int data)
+{
+       struct btrfs_block_group_cache *cache;
+       struct btrfs_fs_info *info = root->fs_info;
+       u64 total = num;
+       u64 old_val;
+       u64 block_in_group;
+       u64 i;
+       int ret;
+
+       while(total) {
+               cache = btrfs_lookup_block_group(info, blocknr);
+               if (!cache) {
+                       return -1;
+               }
+               block_in_group = blocknr - cache->key.objectid;
+               WARN_ON(block_in_group > cache->key.offset);
+               radix_tree_tag_set(cache->radix, cache->key.objectid +
+                                  cache->key.offset - 1,
+                                  BTRFS_BLOCK_GROUP_DIRTY);
+
+               old_val = btrfs_block_group_used(&cache->item);
+               num = min(total, cache->key.offset - block_in_group);
+               if (alloc) {
+                       if (blocknr > cache->last_alloc)
+                               cache->last_alloc = blocknr;
+                       if (!cache->data) {
+                               for (i = 0; i < num; i++) {
+                                       clear_radix_bit(&info->extent_map_radix,
+                                                       blocknr + i);
+                               }
+                       }
+                       if (cache->data != data &&
+                           old_val < (cache->key.offset >> 1)) {
+                               cache->data = data;
+                               radix_tree_delete(cache->radix,
+                                                 cache->key.objectid +
+                                                 cache->key.offset - 1);
+
+                               if (data) {
+                                       cache->radix =
+                                               &info->block_group_data_radix;
+                                       cache->item.flags |=
+                                               BTRFS_BLOCK_GROUP_DATA;
+                               } else {
+                                       cache->radix = &info->block_group_radix;
+                                       cache->item.flags &=
+                                               ~BTRFS_BLOCK_GROUP_DATA;
+                               }
+                               ret = radix_tree_insert(cache->radix,
+                                                       cache->key.objectid +
+                                                       cache->key.offset - 1,
+                                                       (void *)cache);
+                       }
+                       old_val += num;
+               } else {
+                       old_val -= num;
+                       if (blocknr < cache->first_free)
+                               cache->first_free = blocknr;
+                       if (!cache->data && mark_free) {
+                               for (i = 0; i < num; i++) {
+                                       set_radix_bit(&info->extent_map_radix,
+                                                     blocknr + i);
+                               }
+                       }
+                       if (old_val < (cache->key.offset >> 1) &&
+                           old_val + num >= (cache->key.offset >> 1)) {
+                               radix_tree_tag_set(cache->radix,
+                                                  cache->key.objectid +
+                                                  cache->key.offset - 1,
+                                                  BTRFS_BLOCK_GROUP_AVAIL);
+                       }
                }
+               btrfs_set_block_group_used(&cache->item, old_val);
+               total -= num;
+               blocknr += num;
        }
        return 0;
 }
 
-int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
-                              btrfs_root *root)
+int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy)
 {
        unsigned long gang[8];
+       u64 last = 0;
+       struct radix_tree_root *pinned_radix = &root->fs_info->pinned_radix;
+       int ret;
+       int i;
+
+       while(1) {
+               ret = find_first_radix_bit(pinned_radix, gang, last,
+                                          ARRAY_SIZE(gang));
+               if (!ret)
+                       break;
+               for (i = 0 ; i < ret; i++) {
+                       set_radix_bit(copy, gang[i]);
+                       last = gang[i] + 1;
+               }
+       }
+       ret = find_first_radix_bit(&root->fs_info->extent_ins_radix, gang, 0,
+                                  ARRAY_SIZE(gang));
+       WARN_ON(ret);
+       return 0;
+}
+
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
+                              struct btrfs_root *root,
+                              struct radix_tree_root *unpin_radix)
+{
+       unsigned long gang[8];
+       struct btrfs_block_group_cache *block_group;
        u64 first = 0;
        int ret;
        int i;
        struct radix_tree_root *pinned_radix = &root->fs_info->pinned_radix;
+       struct radix_tree_root *extent_radix = &root->fs_info->extent_map_radix;
 
        while(1) {
-               ret = find_first_radix_bit(pinned_radix, gang,
+               ret = find_first_radix_bit(unpin_radix, gang, 0,
                                           ARRAY_SIZE(gang));
                if (!ret)
                        break;
@@ -145,11 +741,19 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
                        first = gang[0];
                for (i = 0; i < ret; i++) {
                        clear_radix_bit(pinned_radix, gang[i]);
+                       clear_radix_bit(unpin_radix, gang[i]);
+                       block_group = btrfs_lookup_block_group(root->fs_info,
+                                                              gang[i]);
+                       if (block_group) {
+                               WARN_ON(block_group->pinned == 0);
+                               block_group->pinned--;
+                               if (gang[i] < block_group->last_alloc)
+                                       block_group->last_alloc = gang[i];
+                               if (!block_group->data)
+                                       set_radix_bit(extent_radix, gang[i]);
+                       }
                }
        }
-       if (root->fs_info->last_insert.objectid > first)
-               root->fs_info->last_insert.objectid = first;
-       root->fs_info->last_insert.offset = 0;
        return 0;
 }
 
@@ -160,25 +764,31 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
        struct btrfs_extent_item extent_item;
        int i;
        int ret;
-       u64 super_blocks_used;
+       int err;
+       unsigned long gang[8];
        struct btrfs_fs_info *info = extent_root->fs_info;
 
        btrfs_set_extent_refs(&extent_item, 1);
        ins.offset = 1;
        ins.flags = 0;
        btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
+       btrfs_set_extent_owner(&extent_item, extent_root->root_key.objectid);
 
-       for (i = 0; i < extent_root->fs_info->current_insert.flags; i++) {
-               ins.objectid = extent_root->fs_info->current_insert.objectid +
-                               i;
-               super_blocks_used = btrfs_super_blocks_used(info->disk_super);
-               btrfs_set_super_blocks_used(info->disk_super,
-                                           super_blocks_used + 1);
-               ret = btrfs_insert_item(trans, extent_root, &ins, &extent_item,
-                                       sizeof(extent_item));
-               BUG_ON(ret);
+       while(1) {
+               ret = find_first_radix_bit(&info->extent_ins_radix, gang, 0,
+                                          ARRAY_SIZE(gang));
+               if (!ret)
+                       break;
+
+               for (i = 0; i < ret; i++) {
+                       ins.objectid = gang[i];
+                       err = btrfs_insert_item(trans, extent_root, &ins,
+                                               &extent_item,
+                                               sizeof(extent_item));
+                       clear_radix_bit(&info->extent_ins_radix, gang[i]);
+                       WARN_ON(err);
+               }
        }
-       extent_root->fs_info->current_insert.offset = 0;
        return 0;
 }
 
@@ -204,10 +814,17 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
                        btrfs_block_release(root, bh);
                }
                err = set_radix_bit(&root->fs_info->pinned_radix, blocknr);
+               if (!err) {
+                       struct btrfs_block_group_cache *cache;
+                       cache = btrfs_lookup_block_group(root->fs_info,
+                                                        blocknr);
+                       if (cache)
+                               cache->pinned++;
+               }
        } else {
                err = set_radix_bit(&root->fs_info->pending_del_radix, blocknr);
        }
-       BUG_ON(err);
+       BUG_ON(err < 0);
        return 0;
 }
 
@@ -215,7 +832,8 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
  * remove an extent from the root, returns 0 on success
  */
 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-                        *root, u64 blocknr, u64 num_blocks, int pin)
+                        *root, u64 blocknr, u64 num_blocks, int pin,
+                        int mark_free)
 {
        struct btrfs_path *path;
        struct btrfs_key key;
@@ -223,7 +841,6 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
        struct btrfs_root *extent_root = info->extent_root;
        int ret;
        struct btrfs_extent_item *ei;
-       struct btrfs_key ins;
        u32 refs;
 
        key.objectid = blocknr;
@@ -231,18 +848,14 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
        btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
        key.offset = num_blocks;
 
-       find_free_extent(trans, root, 0, 0, (u64)-1, &ins);
        path = btrfs_alloc_path();
-       BUG_ON(!path);
-       btrfs_init_path(path);
+       if (!path)
+               return -ENOMEM;
 
        ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
-       if (ret) {
-               printk("failed to find %Lu\n", key.objectid);
-               btrfs_print_tree(extent_root, extent_root->node);
-               printk("failed to find %Lu\n", key.objectid);
-               BUG();
-       }
+       if (ret < 0)
+               return ret;
+       BUG_ON(ret);
        ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
                            struct btrfs_extent_item);
        BUG_ON(ei->refs == 0);
@@ -250,21 +863,31 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
        btrfs_set_extent_refs(ei, refs);
        btrfs_mark_buffer_dirty(path->nodes[0]);
        if (refs == 0) {
-               u64 super_blocks_used;
+               u64 super_blocks_used, root_blocks_used;
 
                if (pin) {
                        ret = pin_down_block(root, blocknr, 0);
                        BUG_ON(ret);
                }
 
-               super_blocks_used = btrfs_super_blocks_used(info->disk_super);
-               btrfs_set_super_blocks_used(info->disk_super,
+               /* block accounting for super block */
+               super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
+               btrfs_set_super_blocks_used(&info->super_copy,
                                            super_blocks_used - num_blocks);
+
+               /* block accounting for root item */
+               root_blocks_used = btrfs_root_blocks_used(&root->root_item);
+               btrfs_set_root_blocks_used(&root->root_item,
+                                          root_blocks_used - num_blocks);
+
                ret = btrfs_del_item(trans, extent_root, path);
-               if (ret)
-                       BUG();
+               if (ret) {
+                       return ret;
+               }
+               ret = update_block_group(trans, root, blocknr, num_blocks, 0,
+                                        mark_free, 0);
+               BUG_ON(ret);
        }
-       btrfs_release_path(extent_root, path);
        btrfs_free_path(path);
        finish_current_insert(trans, extent_root);
        return ret;
@@ -284,22 +907,34 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
        int i;
        struct radix_tree_root *pending_radix;
        struct radix_tree_root *pinned_radix;
+       struct btrfs_block_group_cache *cache;
 
        pending_radix = &extent_root->fs_info->pending_del_radix;
        pinned_radix = &extent_root->fs_info->pinned_radix;
 
        while(1) {
-               ret = find_first_radix_bit(pending_radix, gang,
+               ret = find_first_radix_bit(pending_radix, gang, 0,
                                           ARRAY_SIZE(gang));
                if (!ret)
                        break;
                for (i = 0; i < ret; i++) {
                        wret = set_radix_bit(pinned_radix, gang[i]);
-                       BUG_ON(wret);
+                       if (wret == 0) {
+                               cache =
+                                 btrfs_lookup_block_group(extent_root->fs_info,
+                                                          gang[i]);
+                               if (cache)
+                                       cache->pinned++;
+                       }
+                       if (wret < 0) {
+                               printk(KERN_CRIT "set_radix_bit, err %d\n",
+                                      wret);
+                               BUG_ON(wret < 0);
+                       }
                        wret = clear_radix_bit(pending_radix, gang[i]);
                        BUG_ON(wret);
                        wret = __free_extent(trans, extent_root,
-                                            gang[i], 1, 0);
+                                            gang[i], 1, 0, 0);
                        if (wret)
                                err = wret;
                }
@@ -321,7 +956,7 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
                pin_down_block(root, blocknr, 1);
                return 0;
        }
-       ret = __free_extent(trans, root, blocknr, num_blocks, pin);
+       ret = __free_extent(trans, root, blocknr, num_blocks, pin, pin == 0);
        pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
        return ret ? ret : pending_ret;
 }
@@ -335,8 +970,10 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
  * Any available blocks before search_start are skipped.
  */
 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-                           *orig_root, u64 num_blocks, u64 search_start, u64
-                           search_end, struct btrfs_key *ins)
+                           *orig_root, u64 num_blocks, u64 empty_size,
+                           u64 search_start, u64 search_end, u64 hint_block,
+                           struct btrfs_key *ins, u64 exclude_start,
+                           u64 exclude_nr, int data)
 {
        struct btrfs_path *path;
        struct btrfs_key key;
@@ -345,47 +982,77 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
        int slot = 0;
        u64 last_block = 0;
        u64 test_block;
+       u64 orig_search_start = search_start;
        int start_found;
        struct btrfs_leaf *l;
        struct btrfs_root * root = orig_root->fs_info->extent_root;
+       struct btrfs_fs_info *info = root->fs_info;
        int total_needed = num_blocks;
        int level;
+       struct btrfs_block_group_cache *block_group;
+       int full_scan = 0;
+       int wrapped = 0;
 
-       path = btrfs_alloc_path();
+       WARN_ON(num_blocks < 1);
        ins->flags = 0;
        btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 
        level = btrfs_header_level(btrfs_buffer_header(root->node));
-       total_needed += (level + 1) * 3;
-       if (root->fs_info->last_insert.objectid == 0 && search_end == (u64)-1) {
-               struct btrfs_disk_key *last_key;
-               btrfs_init_path(path);
-               ins->objectid = (u64)-1;
-               ins->offset = (u64)-1;
-               ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
-               if (ret < 0)
-                       goto error;
-               BUG_ON(ret == 0);
-               if (path->slots[0] > 0)
-                       path->slots[0]--;
-               l = btrfs_buffer_leaf(path->nodes[0]);
-               last_key = &l->items[path->slots[0]].key;
-               search_start = btrfs_disk_key_objectid(last_key);
+       if (search_end == (u64)-1)
+               search_end = btrfs_super_total_blocks(&info->super_copy);
+       if (hint_block) {
+               block_group = btrfs_lookup_block_group(info, hint_block);
+               block_group = btrfs_find_block_group(root, block_group,
+                                                    hint_block, data, 1);
+       } else {
+               block_group = btrfs_find_block_group(root,
+                                                    trans->block_group, 0,
+                                                    data, 1);
        }
-       if (root->fs_info->last_insert.objectid > search_start)
-               search_start = root->fs_info->last_insert.objectid;
+
+       total_needed += empty_size;
+       path = btrfs_alloc_path();
 
 check_failed:
+       if (!block_group->data)
+               search_start = find_search_start(root, &block_group,
+                                                search_start, total_needed);
+       else if (!full_scan)
+               search_start = max(block_group->last_alloc, search_start);
+
        btrfs_init_path(path);
        ins->objectid = search_start;
        ins->offset = 0;
        start_found = 0;
+       path->reada = 2;
+
        ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
        if (ret < 0)
                goto error;
 
-       if (path->slots[0] > 0)
+       if (path->slots[0] > 0) {
                path->slots[0]--;
+       }
+
+       l = btrfs_buffer_leaf(path->nodes[0]);
+       btrfs_disk_key_to_cpu(&key, &l->items[path->slots[0]].key);
+       /*
+        * a rare case, go back one key if we hit a block group item
+        * instead of an extent item
+        */
+       if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY &&
+           key.objectid + key.offset >= search_start) {
+               ins->objectid = key.objectid;
+               ins->offset = key.offset - 1;
+               btrfs_release_path(root, path);
+               ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
+               if (ret < 0)
+                       goto error;
+
+               if (path->slots[0] > 0) {
+                       path->slots[0]--;
+               }
+       }
 
        while (1) {
                l = btrfs_buffer_leaf(path->nodes[0]);
@@ -398,61 +1065,114 @@ check_failed:
                                goto error;
                        if (!start_found) {
                                ins->objectid = search_start;
-                               ins->offset = (u64)-1;
+                               ins->offset = search_end - search_start;
                                start_found = 1;
                                goto check_pending;
                        }
                        ins->objectid = last_block > search_start ?
                                        last_block : search_start;
-                       ins->offset = (u64)-1;
+                       ins->offset = search_end - ins->objectid;
                        goto check_pending;
                }
+
                btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
-               if (key.objectid >= search_start) {
-                       if (start_found) {
-                               if (last_block < search_start)
-                                       last_block = search_start;
-                               hole_size = key.objectid - last_block;
-                               if (hole_size > total_needed) {
-                                       ins->objectid = last_block;
-                                       ins->offset = hole_size;
-                                       goto check_pending;
-                               }
+               if (key.objectid >= search_start && key.objectid > last_block &&
+                   start_found) {
+                       if (last_block < search_start)
+                               last_block = search_start;
+                       hole_size = key.objectid - last_block;
+                       if (hole_size >= num_blocks) {
+                               ins->objectid = last_block;
+                               ins->offset = hole_size;
+                               goto check_pending;
                        }
                }
+
+               if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
+                       goto next;
+
                start_found = 1;
                last_block = key.objectid + key.offset;
+               if (!full_scan && last_block >= block_group->key.objectid +
+                   block_group->key.offset) {
+                       btrfs_release_path(root, path);
+                       search_start = block_group->key.objectid +
+                               block_group->key.offset * 2;
+                       goto new_group;
+               }
+next:
                path->slots[0]++;
+               cond_resched();
        }
-       // FIXME -ENOSPC
 check_pending:
        /* we have to make sure we didn't find an extent that has already
         * been allocated by the map tree or the original allocation
         */
        btrfs_release_path(root, path);
        BUG_ON(ins->objectid < search_start);
+
+       if (ins->objectid + num_blocks >= search_end) {
+               if (full_scan) {
+                       ret = -ENOSPC;
+                       goto error;
+               }
+               search_start = orig_search_start;
+               if (wrapped) {
+                       if (!full_scan)
+                               total_needed -= empty_size;
+                       full_scan = 1;
+               } else
+                       wrapped = 1;
+               goto new_group;
+       }
        for (test_block = ins->objectid;
-            test_block < ins->objectid + total_needed; test_block++) {
-               if (test_radix_bit(&root->fs_info->pinned_radix,
-                                     test_block)) {
+            test_block < ins->objectid + num_blocks; test_block++) {
+               if (test_radix_bit(&info->pinned_radix, test_block) ||
+                   test_radix_bit(&info->extent_ins_radix, test_block)) {
                        search_start = test_block + 1;
-                       goto check_failed;
+                       goto new_group;
                }
        }
-       BUG_ON(root->fs_info->current_insert.offset);
-       root->fs_info->current_insert.offset = total_needed - num_blocks;
-       root->fs_info->current_insert.objectid = ins->objectid + num_blocks;
-       root->fs_info->current_insert.flags = 0;
-       root->fs_info->last_insert.objectid = ins->objectid;
+       if (exclude_nr > 0 && (ins->objectid + num_blocks > exclude_start &&
+           ins->objectid < exclude_start + exclude_nr)) {
+               search_start = exclude_start + exclude_nr;
+               goto new_group;
+       }
+       if (!data) {
+               block_group = btrfs_lookup_block_group(info, ins->objectid);
+               if (block_group)
+                       trans->block_group = block_group;
+       }
        ins->offset = num_blocks;
        btrfs_free_path(path);
        return 0;
+
+new_group:
+       if (search_start + num_blocks >= search_end) {
+               search_start = orig_search_start;
+               if (full_scan) {
+                       ret = -ENOSPC;
+                       goto error;
+               }
+               if (wrapped) {
+                       if (!full_scan)
+                               total_needed -= empty_size;
+                       full_scan = 1;
+               } else
+                       wrapped = 1;
+       }
+       block_group = btrfs_lookup_block_group(info, search_start);
+       cond_resched();
+       if (!full_scan)
+               block_group = btrfs_find_block_group(root, block_group,
+                                                    search_start, data, 0);
+       goto check_failed;
+
 error:
        btrfs_release_path(root, path);
        btrfs_free_path(path);
        return ret;
 }
-
 /*
  * finds a free extent and does all the dirty work required for allocation
  * returns the key for the extent through ins, and a tree buffer for
@@ -462,47 +1182,68 @@ error:
  */
 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
                       struct btrfs_root *root, u64 owner,
-                      u8 type, u64 num_blocks, u64 search_start,
-                      u64 search_end, struct btrfs_key *ins)
+                      u64 num_blocks, u64 empty_size, u64 hint_block,
+                      u64 search_end, struct btrfs_key *ins, int data)
 {
        int ret;
        int pending_ret;
-       u64 super_blocks_used;
+       u64 super_blocks_used, root_blocks_used;
+       u64 search_start = 0;
        struct btrfs_fs_info *info = root->fs_info;
        struct btrfs_root *extent_root = info->extent_root;
        struct btrfs_extent_item extent_item;
 
        btrfs_set_extent_refs(&extent_item, 1);
        btrfs_set_extent_owner(&extent_item, owner);
-       btrfs_set_extent_type(&extent_item, type);
 
-       if (root == extent_root) {
-               BUG_ON(extent_root->fs_info->current_insert.offset == 0);
-               BUG_ON(num_blocks != 1);
-               BUG_ON(extent_root->fs_info->current_insert.flags ==
-                      extent_root->fs_info->current_insert.offset);
-               ins->offset = 1;
-               ins->objectid = extent_root->fs_info->current_insert.objectid +
-                               extent_root->fs_info->current_insert.flags++;
-               return 0;
-       }
-       ret = find_free_extent(trans, root, num_blocks, search_start,
-                              search_end, ins);
+       WARN_ON(num_blocks < 1);
+       ret = find_free_extent(trans, root, num_blocks, empty_size,
+                              search_start, search_end, hint_block, ins,
+                              trans->alloc_exclude_start,
+                              trans->alloc_exclude_nr, data);
+       BUG_ON(ret);
        if (ret)
                return ret;
 
-       super_blocks_used = btrfs_super_blocks_used(info->disk_super);
-       btrfs_set_super_blocks_used(info->disk_super, super_blocks_used +
+       /* block accounting for super block */
+       super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
+       btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used +
                                    num_blocks);
+
+       /* block accounting for root item */
+       root_blocks_used = btrfs_root_blocks_used(&root->root_item);
+       btrfs_set_root_blocks_used(&root->root_item, root_blocks_used +
+                                  num_blocks);
+
+       if (root == extent_root) {
+               BUG_ON(num_blocks != 1);
+               set_radix_bit(&root->fs_info->extent_ins_radix, ins->objectid);
+               goto update_block;
+       }
+
+       WARN_ON(trans->alloc_exclude_nr);
+       trans->alloc_exclude_start = ins->objectid;
+       trans->alloc_exclude_nr = ins->offset;
        ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
                                sizeof(extent_item));
 
+       trans->alloc_exclude_start = 0;
+       trans->alloc_exclude_nr = 0;
+
+       BUG_ON(ret);
        finish_current_insert(trans, extent_root);
        pending_ret = del_pending_extents(trans, extent_root);
-       if (ret)
+       if (ret) {
                return ret;
-       if (pending_ret)
+       }
+       if (pending_ret) {
                return pending_ret;
+       }
+
+update_block:
+       ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0,
+                                data);
+       BUG_ON(ret);
        return 0;
 }
 
@@ -511,21 +1252,30 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
  * returns the tree buffer or NULL.
  */
 struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-                                          struct btrfs_root *root)
+                                          struct btrfs_root *root, u64 hint,
+                                          u64 empty_size)
 {
        struct btrfs_key ins;
        int ret;
        struct buffer_head *buf;
 
        ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
-                                BTRFS_EXTENT_TREE,
-                                1, 0, (unsigned long)-1, &ins);
+                                1, empty_size, hint, (u64)-1, &ins, 0);
        if (ret) {
-               BUG();
-               return NULL;
+               BUG_ON(ret > 0);
+               return ERR_PTR(ret);
        }
        buf = btrfs_find_create_tree_block(root, ins.objectid);
+       if (!buf) {
+               btrfs_free_extent(trans, root, ins.objectid, 1, 0);
+               return ERR_PTR(-ENOMEM);
+       }
+       WARN_ON(buffer_dirty(buf));
        set_buffer_uptodate(buf);
+       set_buffer_checked(buf);
+       set_buffer_defrag(buf);
+       set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
+       trans->blocks_used++;
        return buf;
 }
 
@@ -543,6 +1293,7 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
        leaf = btrfs_buffer_leaf(cur);
        nritems = btrfs_header_nritems(&leaf->header);
        for (i = 0; i < nritems; i++) {
+               u64 disk_blocknr;
                key = &leaf->items[i].key;
                if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
                        continue;
@@ -553,8 +1304,10 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
                 * FIXME make sure to insert a trans record that
                 * repeats the snapshot del on crash
                 */
-               ret = btrfs_free_extent(trans, root,
-                                       btrfs_file_extent_disk_blocknr(fi),
+               disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+               if (disk_blocknr == 0)
+                       continue;
+               ret = btrfs_free_extent(trans, root, disk_blocknr,
                                        btrfs_file_extent_disk_num_blocks(fi),
                                        0);
                BUG_ON(ret);
@@ -562,6 +1315,31 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
        return 0;
 }
 
+static void reada_walk_down(struct btrfs_root *root,
+                           struct btrfs_node *node)
+{
+       int i;
+       u32 nritems;
+       u64 blocknr;
+       int ret;
+       u32 refs;
+
+       nritems = btrfs_header_nritems(&node->header);
+       for (i = 0; i < nritems; i++) {
+               blocknr = btrfs_node_blockptr(node, i);
+               ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
+               BUG_ON(ret);
+               if (refs != 1)
+                       continue;
+               mutex_unlock(&root->fs_info->fs_mutex);
+               ret = readahead_tree_block(root, blocknr);
+               cond_resched();
+               mutex_lock(&root->fs_info->fs_mutex);
+               if (ret)
+                       break;
+       }
+}
+
 /*
  * helper function for drop_snapshot, this walks down the tree dropping ref
  * counts as it goes.
@@ -582,6 +1360,7 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
        BUG_ON(ret);
        if (refs > 1)
                goto out;
+
        /*
         * walk down to the last node level and free all the leaves
         */
@@ -589,8 +1368,13 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
                WARN_ON(*level < 0);
                WARN_ON(*level >= BTRFS_MAX_LEVEL);
                cur = path->nodes[*level];
+
+               if (*level > 0 && path->slots[*level] == 0)
+                       reada_walk_down(root, btrfs_buffer_node(cur));
+
                if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
                        WARN_ON(1);
+
                if (path->slots[*level] >=
                    btrfs_header_nritems(btrfs_buffer_header(cur)))
                        break;
@@ -609,7 +1393,25 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
                        BUG_ON(ret);
                        continue;
                }
-               next = read_tree_block(root, blocknr);
+               next = btrfs_find_tree_block(root, blocknr);
+               if (!next || !buffer_uptodate(next)) {
+                       brelse(next);
+                       mutex_unlock(&root->fs_info->fs_mutex);
+                       next = read_tree_block(root, blocknr);
+                       mutex_lock(&root->fs_info->fs_mutex);
+
+                       /* we dropped the lock, check one more time */
+                       ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
+                       BUG_ON(ret);
+                       if (refs != 1) {
+                               path->slots[*level]++;
+                               brelse(next);
+                               ret = btrfs_free_extent(trans, root,
+                                                       blocknr, 1, 1);
+                               BUG_ON(ret);
+                               continue;
+                       }
+               }
                WARN_ON(*level <= 0);
                if (path->nodes[*level-1])
                        btrfs_block_release(root, path->nodes[*level-1]);
@@ -640,12 +1442,21 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
        int i;
        int slot;
        int ret;
+       struct btrfs_root_item *root_item = &root->root_item;
+
        for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
                slot = path->slots[i];
                if (slot < btrfs_header_nritems(
                    btrfs_buffer_header(path->nodes[i])) - 1) {
+                       struct btrfs_node *node;
+                       node = btrfs_buffer_node(path->nodes[i]);
                        path->slots[i]++;
                        *level = i;
+                       WARN_ON(*level == 0);
+                       memcpy(&root_item->drop_progress,
+                              &node->ptrs[path->slots[i]].key,
+                              sizeof(root_item->drop_progress));
+                       root_item->drop_level = i;
                        return 0;
                } else {
                        ret = btrfs_free_extent(trans, root,
@@ -666,7 +1477,7 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
  * decremented.
  */
 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
-                       *root, struct buffer_head *snap)
+                       *root)
 {
        int ret = 0;
        int wret;
@@ -674,15 +1485,34 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
        struct btrfs_path *path;
        int i;
        int orig_level;
+       struct btrfs_root_item *root_item = &root->root_item;
 
        path = btrfs_alloc_path();
        BUG_ON(!path);
-       btrfs_init_path(path);
 
-       level = btrfs_header_level(btrfs_buffer_header(snap));
+       level = btrfs_header_level(btrfs_buffer_header(root->node));
        orig_level = level;
-       path->nodes[level] = snap;
-       path->slots[level] = 0;
+       if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
+               path->nodes[level] = root->node;
+               path->slots[level] = 0;
+       } else {
+               struct btrfs_key key;
+               struct btrfs_disk_key *found_key;
+               struct btrfs_node *node;
+
+               btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
+               level = root_item->drop_level;
+               path->lowest_level = level;
+               wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+               if (wret < 0) {
+                       ret = wret;
+                       goto out;
+               }
+               node = btrfs_buffer_node(path->nodes[level]);
+               found_key = &node->ptrs[path->slots[level]].key;
+               WARN_ON(memcmp(found_key, &root_item->drop_progress,
+                              sizeof(*found_key)));
+       }
        while(1) {
                wret = walk_down_tree(trans, root, path, &level);
                if (wret > 0)
@@ -695,12 +1525,146 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
                        break;
                if (wret < 0)
                        ret = wret;
+               ret = -EAGAIN;
+               get_bh(root->node);
+               break;
        }
        for (i = 0; i <= orig_level; i++) {
                if (path->nodes[i]) {
                        btrfs_block_release(root, path->nodes[i]);
+                       path->nodes[i] = 0;
                }
        }
+out:
        btrfs_free_path(path);
        return ret;
 }
+
+static int free_block_group_radix(struct radix_tree_root *radix)
+{
+       int ret;
+       struct btrfs_block_group_cache *cache[8];
+       int i;
+
+       while(1) {
+               ret = radix_tree_gang_lookup(radix, (void **)cache, 0,
+                                            ARRAY_SIZE(cache));
+               if (!ret)
+                       break;
+               for (i = 0; i < ret; i++) {
+                       radix_tree_delete(radix, cache[i]->key.objectid +
+                                         cache[i]->key.offset - 1);
+                       kfree(cache[i]);
+               }
+       }
+       return 0;
+}
+
+int btrfs_free_block_groups(struct btrfs_fs_info *info)
+{
+       int ret;
+       int ret2;
+       unsigned long gang[16];
+       int i;
+
+       ret = free_block_group_radix(&info->block_group_radix);
+       ret2 = free_block_group_radix(&info->block_group_data_radix);
+       if (ret)
+               return ret;
+       if (ret2)
+               return ret2;
+
+       while(1) {
+               ret = find_first_radix_bit(&info->extent_map_radix,
+                                          gang, 0, ARRAY_SIZE(gang));
+               if (!ret)
+                       break;
+               for (i = 0; i < ret; i++) {
+                       clear_radix_bit(&info->extent_map_radix, gang[i]);
+               }
+       }
+       return 0;
+}
+
+int btrfs_read_block_groups(struct btrfs_root *root)
+{
+       struct btrfs_path *path;
+       int ret;
+       int err = 0;
+       struct btrfs_block_group_item *bi;
+       struct btrfs_block_group_cache *cache;
+       struct btrfs_fs_info *info = root->fs_info;
+       struct radix_tree_root *radix;
+       struct btrfs_key key;
+       struct btrfs_key found_key;
+       struct btrfs_leaf *leaf;
+       u64 group_size_blocks;
+       u64 used;
+
+       group_size_blocks = BTRFS_BLOCK_GROUP_SIZE >>
+               root->fs_info->sb->s_blocksize_bits;
+       root = info->extent_root;
+       key.objectid = 0;
+       key.offset = group_size_blocks;
+       key.flags = 0;
+       btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return -ENOMEM;
+
+       while(1) {
+               ret = btrfs_search_slot(NULL, info->extent_root,
+                                       &key, path, 0, 0);
+               if (ret != 0) {
+                       err = ret;
+                       break;
+               }
+               leaf = btrfs_buffer_leaf(path->nodes[0]);
+               btrfs_disk_key_to_cpu(&found_key,
+                                     &leaf->items[path->slots[0]].key);
+               cache = kmalloc(sizeof(*cache), GFP_NOFS);
+               if (!cache) {
+                       err = -1;
+                       break;
+               }
+
+               bi = btrfs_item_ptr(leaf, path->slots[0],
+                                   struct btrfs_block_group_item);
+               if (bi->flags & BTRFS_BLOCK_GROUP_DATA) {
+                       radix = &info->block_group_data_radix;
+                       cache->data = 1;
+               } else {
+                       radix = &info->block_group_radix;
+                       cache->data = 0;
+               }
+
+               memcpy(&cache->item, bi, sizeof(*bi));
+               memcpy(&cache->key, &found_key, sizeof(found_key));
+               cache->last_alloc = cache->key.objectid;
+               cache->first_free = cache->key.objectid;
+               cache->pinned = 0;
+               cache->cached = 0;
+
+               cache->radix = radix;
+
+               key.objectid = found_key.objectid + found_key.offset;
+               btrfs_release_path(root, path);
+               ret = radix_tree_insert(radix, found_key.objectid +
+                                       found_key.offset - 1,
+                                       (void *)cache);
+               BUG_ON(ret);
+               used = btrfs_block_group_used(bi);
+               if (used < div_factor(key.offset, 8)) {
+                       radix_tree_tag_set(radix, found_key.objectid +
+                                          found_key.offset - 1,
+                                          BTRFS_BLOCK_GROUP_AVAIL);
+               }
+               if (key.objectid >=
+                   btrfs_super_total_blocks(&info->super_copy))
+                       break;
+       }
+
+       btrfs_free_path(path);
+       return 0;
+}