From glancing at the source code I believe this uses the same algorithm as df from coreutils. It outputs pretty much useless data for btrfs, which needs to be h

From glancing at the source code I believe this uses the same algorithm as df from coreutils. It outputs pretty much useless data for btrfs, which needs to be handled differently. See: $ btrfs fi df $ btrfs fi us