So the problem with the last patch was that it did not handle blocksize < pagesize. On a x86 box, with a 2K/4K file system, it was only initializing the first 2K block. This fixes it. Sunil