Il 03/12/2013 16:09, Kevin Wolf ha scritto: >> > BDRVQcowState *s = bs->opaque; >> > + bdi->unallocated_blocks_are_zero = true; >> > + bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3); >> > bdi->cluster_size = s->cluster_size; >> > bdi->vm_state_offset = qcow2_vm_state_offset(s); >> > return 0; > We must change qcow2_discard_clusters() to set the zero flag instead of > just deallocating the cluster. (We should be doing that anyway, because > the backing file reappearing isn't very nice.)
No, that's not needed: * unallocated_blocks_are_zero is only meaningful for bs->backing_hd == NULL (not too intuitive, but I didn't introduce that interface :)). In fact, v2 was checking backing_hd == NULL but I removed it after Peter noticed I was being inconsistent. * can_write_zeroes_with_unmap correctly returns true only if zero clusters are enabled > For the other formats, I guess this is only correct because they don't > implement discard anyway? No, it is correct because that's what their bdrv_co_readv (or similar) will return when a block is not allocated and there is no backing file. Of course, for many formats there will never be a backing file. Paolo
