Hi! I wanted to experiment with sector size = 4096, to see how an OS reacts. And naturally, the first instinct is to try it in qemu. But it turned to be.. difficult.
First, there's no documentation. But I remember seeing properties of ide-hd device, like physical_sector_size etc. Ok. $ qemu-img create -f raw img 32G $ kvm -machine q35 -drive file=img,format=raw,if=none,id=d \ -device ide-hd,drive=d,physical_sector_size=4096 This way, the guest see size of the drive as 0. This is obviously not right (tm). Okay, lets try full 4096 sectors: -device ide-hd,drive=d,physical_sector_size=4096,logical_sector_size=4096 This fails: discard_granularity must be a multiple of logical_block_size Let's adjust discard_granularity (I think it'd be good if qemu does this based on physical_sector_size already): -device ide-hd,drive=d,physical_sector_size=4096,logical_sector_size=4096,discard_granularity=0 This fails: logical_block_size must be 512 for IDE Huh? Should the tell all the sata Advanced-Format (AF) drives manufacturers their drives does not exist, or what? :) Ok. Tried the same with nvme. There, sector sizes seems to be entirely ignored, - no matter which values I choose, guest always see them as 512. What's the proper way to run guest with 4096-byte sector size in qemu? Thanks, /mjt
