This helps prevent tmpfs dirtiers from skewing the per-cpu bdp_ratelimits.
Acked-by: NJan Kara <jack@suse.cz>
Signed-off-by: NWu Fengguang <fengguang.wu@intel.com>