On 05.06.20 15:00, Kevin Wolf wrote: [...]
> Makes sense. I'm not against preallocation during image creation. I just > think it shouldn't play a role in deciding whether an image is valid or > not. Oh, no. I wouldn’t consider it corrupted just because some clusters are not allocated. I’d just say any program that handles qcow2 files is responsible to ensure that images with the data-file-raw flag actually fulfill the flag’s promise. So if a cluster isn’t allocated in qcow2, it must read as zeroes in the data file (because the spec disallows backing files with data-file-raw anyway[1]). It’s just my impression that qemu currently doesn’t always ensure this, and the easiest way to do so would be to enforce metadata preallocation for such images. Max [1] Such images are indeed corrupt, hence this patch here from Berto.
signature.asc
Description: OpenPGP digital signature
