This is to reduce the batch size of fstrim to avoid long latency.
Signed-off-by: NJaegeuk Kim <jaegeuk@kernel.org>