On Fri, 09/20 17:04, Paolo Bonzini wrote: > Il 19/09/2013 13:31, Stefan Hajnoczi ha scritto: > >> > We use the extent size as cluster size for flat extents (where no L1/L2 > >> > table is allocated so it's safe). > > Why is the extent size passed as the cluster size parameter? > > I think it's so that the flat extent doesn't take up too many cache entries. > > Paolo
Flat extent doesn't take cache entry at all. It's passed as this because so that more the code path for finding data offset can be the same with sparse: flat is a special case of sparse with only 1 cluster. Otherwize flat needs to be treated specially. Another fix would be pass zero to cluster size parameter but initialize it to extent size after parameter checking. But not quite different. Fam
