AcoSail/kernel_optimize_test
8
0
Fork 0
forked from luck/tmp_suning_uos_patched
Code Pull Requests Activity

Btrfs: reduce lock contention during extent insertion

Browse Source 
We're spending huge amounts of time on lock contention during
end_io processing because we unconditionally assume we are overwriting
an existing extent in the file for each IO.

This checks to see if we are outside i_size, and if so, it uses a
less expensive readonly search of the btree to look for existing
extents.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
This commit is contained in:
Chris Mason 2012-04-27 14:31:29 -04:00
parent fede766f28
commit dc7fdde39e
1 changed files with 7 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
fs/btrfs/file.c
View File

@ -567,6 +567,7 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
int extent_type;
int recow;
int ret;
int modify_tree = -1;
if (drop_cache)
btrfs_drop_extent_cache(inode, start, end - 1, 0);
@ -575,10 +576,13 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
if (!path)
return -ENOMEM;
if (start >= BTRFS_I(inode)->disk_i_size)
modify_tree = 0;
while (1) {
recow = 0;
ret = btrfs_lookup_file_extent(trans, root, path, ino,
search_start, -1);
search_start, modify_tree);
if (ret < 0)
break;
if (ret > 0 && path->slots[0] > 0 && search_start == start) {
@ -634,7 +638,8 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
}
search_start = max(key.offset, start);
if (recow) {
if (recow || !modify_tree) {
modify_tree = -1;
btrfs_release_path(path);
continue;
}