Btrfs: Split the extent_map code into two parts
There is now extent_map for mapping offsets in the file to disk and
extent_io for state tracking, IO submission and extent_bufers.
The new extent_map code shifts from [start,end] pairs to [start,len], and
pushes the locking out into the caller. This allows a few performance
optimizations and is easier to use.
A number of extent_map usage bugs were fixed, mostly with failing
to remove extent_map entries when changing the file.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 1cd8c90..c5bb00f 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -233,8 +233,7 @@
int err = 0;
int i;
struct inode *inode = fdentry(file)->d_inode;
- struct extent_map *em;
- struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
u64 hint_byte;
u64 num_bytes;
u64 start_pos;
@@ -242,11 +241,6 @@
u64 end_pos = pos + write_bytes;
u64 inline_size;
loff_t isize = i_size_read(inode);
- em = alloc_extent_map(GFP_NOFS);
- if (!em)
- return -ENOMEM;
-
- em->bdev = inode->i_sb->s_bdev;
start_pos = pos & ~((u64)root->sectorsize - 1);
num_bytes = (write_bytes + pos - start_pos +
@@ -254,7 +248,7 @@
end_of_last_block = start_pos + num_bytes - 1;
- lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS);
+ lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
if (!trans) {
@@ -268,7 +262,7 @@
if ((end_of_last_block & 4095) == 0) {
printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block);
}
- set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS);
+ set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS);
/* FIXME...EIEIO, ENOSPC and more */
@@ -293,6 +287,8 @@
inode->i_ino,
last_pos_in_file,
0, 0, hole_size);
+ btrfs_drop_extent_cache(inode, last_pos_in_file,
+ last_pos_in_file + hole_size -1);
btrfs_check_file(root, inode);
}
if (err)
@@ -320,12 +316,12 @@
last_end += PAGE_CACHE_SIZE - 1;
if (start_pos < isize) {
u64 delalloc_start = start_pos;
- existing_delalloc = count_range_bits(em_tree,
+ existing_delalloc = count_range_bits(io_tree,
&delalloc_start,
end_of_last_block, (u64)-1,
EXTENT_DELALLOC);
}
- set_extent_delalloc(em_tree, start_pos, end_of_last_block,
+ set_extent_delalloc(io_tree, start_pos, end_of_last_block,
GFP_NOFS);
spin_lock(&root->fs_info->delalloc_lock);
root->fs_info->delalloc_bytes += (end_of_last_block + 1 -
@@ -346,6 +342,7 @@
inline_size -= start_pos;
err = insert_inline_extent(trans, root, inode, start_pos,
inline_size, pages, 0, num_pages);
+ btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1);
BUG_ON(err);
}
if (end_pos > isize) {
@@ -356,8 +353,7 @@
err = btrfs_end_transaction(trans, root);
out_unlock:
mutex_unlock(&root->fs_info->fs_mutex);
- unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS);
- free_extent_map(em);
+ unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
return err;
}
@@ -367,10 +363,15 @@
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
while(1) {
+ spin_lock(&em_tree->lock);
em = lookup_extent_mapping(em_tree, start, end);
- if (!em)
+ if (!em) {
+ spin_unlock(&em_tree->lock);
break;
+ }
remove_extent_mapping(em_tree, em);
+ spin_unlock(&em_tree->lock);
+
/* once for us */
free_extent_map(em);
/* once for the tree*/