Btrfs: use a btree inode instead of sb_getblk
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1897f3a..b2ebed7 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -233,6 +233,7 @@
struct btrfs_super_block *disk_super;
struct buffer_head *sb_buffer;
struct super_block *sb;
+ struct inode *btree_inode;
struct mutex trans_mutex;
struct mutex fs_mutex;
};
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 758a62a..970103f 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1,14 +1,17 @@
#include <linux/module.h>
#include <linux/fs.h>
+#include <linux/blkdev.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
+
static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
{
struct btrfs_node *node = btrfs_buffer_node(buf);
- if (buf->b_blocknr != btrfs_header_blocknr(&node->header))
+ if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
BUG();
+ }
if (root->node && btrfs_header_parentid(&node->header) !=
btrfs_header_parentid(btrfs_buffer_header(root->node))) {
BUG();
@@ -16,25 +19,154 @@
return 0;
}
-struct buffer_head *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
+struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
{
- return sb_getblk(root->fs_info->sb, blocknr);
+ struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+ int blockbits = root->fs_info->sb->s_blocksize_bits;
+ unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
+ struct page *page;
+ struct buffer_head *bh;
+ struct buffer_head *head;
+ struct buffer_head *ret = NULL;
+
+ page = find_lock_page(mapping, index);
+ if (!page)
+ return NULL;
+
+ if (!page_has_buffers(page))
+ goto out_unlock;
+
+ head = page_buffers(page);
+ bh = head;
+ do {
+ if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
+ ret = bh;
+ get_bh(bh);
+ goto out_unlock;
+ }
+ bh = bh->b_this_page;
+ } while (bh != head);
+out_unlock:
+ unlock_page(page);
+ page_cache_release(page);
+ return ret;
}
-struct buffer_head *find_tree_block(struct btrfs_root *root, u64 blocknr)
+struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
+ u64 blocknr)
{
- return sb_getblk(root->fs_info->sb, blocknr);
+ struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+ int blockbits = root->fs_info->sb->s_blocksize_bits;
+ unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
+ struct page *page;
+ struct buffer_head *bh;
+ struct buffer_head *head;
+ struct buffer_head *ret = NULL;
+ u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
+ page = grab_cache_page(mapping, index);
+ if (!page)
+ return NULL;
+
+ wait_on_page_writeback(page);
+ if (!page_has_buffers(page))
+ create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
+ head = page_buffers(page);
+ bh = head;
+ do {
+ if (!buffer_mapped(bh)) {
+ bh->b_bdev = root->fs_info->sb->s_bdev;
+ bh->b_blocknr = first_block;
+ set_buffer_mapped(bh);
+ }
+ if (bh->b_blocknr == blocknr) {
+ ret = bh;
+ get_bh(bh);
+ goto out_unlock;
+ }
+ bh = bh->b_this_page;
+ first_block++;
+ } while (bh != head);
+out_unlock:
+ unlock_page(page);
+ page_cache_release(page);
+ return ret;
}
+static sector_t max_block(struct block_device *bdev)
+{
+ sector_t retval = ~((sector_t)0);
+ loff_t sz = i_size_read(bdev->bd_inode);
+
+ if (sz) {
+ unsigned int size = block_size(bdev);
+ unsigned int sizebits = blksize_bits(size);
+ retval = (sz >> sizebits);
+ }
+ return retval;
+}
+
+static int btree_get_block(struct inode *inode, sector_t iblock,
+ struct buffer_head *bh, int create)
+{
+ if (iblock >= max_block(inode->i_sb->s_bdev)) {
+ if (create)
+ return -EIO;
+
+ /*
+ * for reads, we're just trying to fill a partial page.
+ * return a hole, they will have to call get_block again
+ * before they can fill it, and they will get -EIO at that
+ * time
+ */
+ return 0;
+ }
+ bh->b_bdev = inode->i_sb->s_bdev;
+ bh->b_blocknr = iblock;
+ set_buffer_mapped(bh);
+ return 0;
+}
+
+static int btree_writepage(struct page *page, struct writeback_control *wbc)
+{
+ return block_write_full_page(page, btree_get_block, wbc);
+}
+
+static int btree_readpage(struct file * file, struct page * page)
+{
+ return block_read_full_page(page, btree_get_block);
+}
+
+static struct address_space_operations btree_aops = {
+ .readpage = btree_readpage,
+ .writepage = btree_writepage,
+ .sync_page = block_sync_page,
+};
+
struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
{
- struct buffer_head *buf = sb_bread(root->fs_info->sb, blocknr);
+ struct buffer_head *bh = NULL;
- if (!buf)
- return buf;
- if (check_tree_block(root, buf))
+ bh = btrfs_find_create_tree_block(root, blocknr);
+ if (!bh)
+ return bh;
+ lock_buffer(bh);
+ if (!buffer_uptodate(bh)) {
+ get_bh(bh);
+ bh->b_end_io = end_buffer_read_sync;
+ submit_bh(READ, bh);
+ wait_on_buffer(bh);
+ if (!buffer_uptodate(bh))
+ goto fail;
+ } else {
+ unlock_buffer(bh);
+ }
+ if (check_tree_block(root, bh))
BUG();
- return buf;
+ return bh;
+fail:
+ brelse(bh);
+ return NULL;
+
}
int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -101,11 +233,11 @@
GFP_NOFS);
int ret;
- /* FIXME: don't be stupid */
if (!btrfs_super_root(disk_super))
return NULL;
init_bit_radix(&fs_info->pinned_radix);
init_bit_radix(&fs_info->pending_del_radix);
+ sb_set_blocksize(sb, sb_buffer->b_size);
fs_info->running_transaction = NULL;
fs_info->fs_root = root;
fs_info->tree_root = tree_root;
@@ -114,14 +246,30 @@
fs_info->last_inode_alloc = 0;
fs_info->last_inode_alloc_dirid = 0;
fs_info->disk_super = disk_super;
- fs_info->sb_buffer = sb_buffer;
fs_info->sb = sb;
+ fs_info->btree_inode = new_inode(sb);
+ fs_info->btree_inode->i_ino = 1;
+ fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
+ fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
+ mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
+
mutex_init(&fs_info->trans_mutex);
mutex_init(&fs_info->fs_mutex);
memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
__setup_root(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
+
+ fs_info->sb_buffer = read_tree_block(tree_root, sb_buffer->b_blocknr);
+
+ if (!fs_info->sb_buffer)
+ return NULL;
+
+ brelse(sb_buffer);
+ sb_buffer = NULL;
+ disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
+ fs_info->disk_super = disk_super;
+
tree_root->node = read_tree_block(tree_root,
btrfs_super_root(disk_super));
BUG_ON(!tree_root->node);
@@ -137,7 +285,6 @@
ret = find_and_setup_root(disk_super, tree_root, fs_info,
BTRFS_FS_TREE_OBJECTID, root);
BUG_ON(ret);
-
root->commit_root = root->node;
get_bh(root->node);
root->ref_cows = 1;
@@ -191,6 +338,7 @@
root->fs_info->tree_root->node);
btrfs_block_release(root, root->commit_root);
btrfs_block_release(root, root->fs_info->sb_buffer);
+ iput(root->fs_info->btree_inode);
kfree(root->fs_info->extent_root);
kfree(root->fs_info->inode_root);
kfree(root->fs_info->tree_root);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 099f7ee..c2c38bd 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -21,7 +21,8 @@
}
struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr);
-struct buffer_head *find_tree_block(struct btrfs_root *root, u64 blocknr);
+struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
+ u64 blocknr);
int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
struct buffer_head *buf);
int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -37,5 +38,5 @@
void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
int write_ctree_super(struct btrfs_trans_handle *trans,
struct btrfs_root *root);
-int mkfs(int fd, u64 num_blocks, u32 blocksize);
+struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
#endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 176c6dc..b2ae8e7 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -172,8 +172,8 @@
struct buffer_head *bh;
if (!pending) {
- bh = sb_find_get_block(root->fs_info->sb, blocknr);
- if (bh) {
+ bh = btrfs_find_tree_block(root, blocknr);
+ if (bh && buffer_uptodate(bh)) {
header = btrfs_buffer_header(bh);
if (btrfs_header_generation(header) ==
root->fs_info->running_transaction->transid) {
@@ -291,12 +291,10 @@
*root, u64 blocknr, u64 num_blocks, int pin)
{
struct btrfs_root *extent_root = root->fs_info->extent_root;
- struct buffer_head *t;
int pending_ret;
int ret;
if (root == extent_root) {
- t = find_tree_block(root, blocknr);
pin_down_block(root, blocknr, 1);
return 0;
}
@@ -482,7 +480,7 @@
BUG();
return NULL;
}
- buf = find_tree_block(root, ins.objectid);
+ buf = btrfs_find_create_tree_block(root, ins.objectid);
set_buffer_uptodate(buf);
return buf;
}
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 67659b6..8f07f46 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -474,6 +474,8 @@
disk_super = (struct btrfs_super_block *)bh->b_data;
root = open_ctree(sb, bh, disk_super);
sb->s_fs_info = root;
+ disk_super = root->fs_info->disk_super;
+
if (!root) {
printk("btrfs: open_ctree failed\n");
return -EIO;
@@ -734,15 +736,15 @@
struct btrfs_trans_handle *trans;
struct btrfs_root *root;
int ret;
+ root = btrfs_sb(sb);
sb->s_dirt = 0;
if (!wait) {
- filemap_flush(sb->s_bdev->bd_inode->i_mapping);
+ filemap_flush(root->fs_info->btree_inode->i_mapping);
return 0;
}
- filemap_write_and_wait(sb->s_bdev->bd_inode->i_mapping);
+ filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
- root = btrfs_sb(sb);
mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
ret = btrfs_commit_transaction(trans, root);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 46a596e..b20fb53 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -74,7 +74,7 @@
int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
- filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
+ filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
return 0;
}