On Fri, Nov 18, 2011 at 6:29 PM, Kevin Wolf <kw...@redhat.com> wrote: > + /* > + * Increase the refcounts of all clusters and make sure everything is > + * stable on disk before updating the snapshot table to contain a pointer > + * to the new L1 table. > + */ > + ret = qcow2_update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, > 1); > + if (ret < 0) { > + goto fail; > + } > + > + ret = bdrv_flush(bs->file);
Do we need to explicitly flush the qcow2 cache to ensure metadata reaches the disk? Stefan