Comment 10 for bug 1846427

My qcow2 images also reside on an ext4 with features "has_journal ext_attr dir_index filetype needs_recovery extent 64bit flex_bg sparse_super large_file dir_nlink extra_isize metadata_csum" on a luks-encrypt(ed|ing) device mapper device backed by a partition on an NVMe SSD. The setup is rock solid and I had no other indications of it causing corruption or being corrupted.

I did a quick test with a 32GB USB3 flash drive formatted as a super floppy (without partitions nor encryption) as XFS and saw the same corruption though less heavily so, likely because the drive is much slower (~ 60MB/s write instead of ~600MB/s write for the NVMe SSD).

The savevm/loadvm cycle was basically the first reliable and fast reproducer I was able to find. I have a dim recollection that some of my corruptions also did not involve any loadvm/savevm but were much rarer and not as easily reproducible.