On Wed 06 May 2020 05:02:52 PM CEST, Alberto Garcia wrote: >>> - With version 2, this is always 0. >>> + With version 2 or with extended L2 entries (see the >>> next >>> + section), this is always 0. >> >> In your cover letter, you said you changed things to tolerate this >> bit being set even with extended L2 entries. Does this sentence need >> a tweak? > > Not a bad idea.
I just had a look at what happens with v2 images. There the bit is actually checked and the image is marked corrupt, but that only seems to happen in qcow2_get_host_offset(). You can write to a cluster that has that bit set and QEMU won't complain (the bit is however cleared). 'qemu-img check' does not detect any inconsistency either. As I said in my cover letter maybe it's worth preparing a separate series that adds a QCOW2_CLUSTER_INVALID type and includes all these cases. Berto
