On 2/25/2024 9:08 PM, Peter Xu wrote: > On Thu, Feb 22, 2024 at 09:28:36AM -0800, Steve Sistare wrote: >> When migration for cpr is initiated, stop the vm and set state >> RUN_STATE_FINISH_MIGRATE before ram is saved. This eliminates the >> possibility of ram and device state being out of sync, and guarantees >> that a guest in the suspended state remains suspended, because qmp_cont >> rejects a cont command in the RUN_STATE_FINISH_MIGRATE state. >> >> Signed-off-by: Steve Sistare <steven.sist...@oracle.com> > > Reviewed-by: Peter Xu <pet...@redhat.com> > > cpr-reboot mode keeps changing behavior. > > Could we declare it "experimental" until it's solid? Maybe a patch to > document this? > > Normally IMHO we shouldn't merge a feature if it's not complete, however > cpr-reboot is so special that the mode itself is already merged in 8.2 > before I started to merge patches, and it keeps changing things. I don't > know what else we can do here besides declaring it experimental and not > declare it a stable feature.
Hi Peter, the planned/committed functionality for cpr-reboot changed only once, in: migration: stop vm for cpr Suspension to support vfio is an enhancement which adds to the basic functionality, it does not change it. This was planned all along, but submitted as a separate series to manage complexity, as I outlined in my qemu community presentation, which I emailed you at the time. Other "changes" that arose during review were just clarifications and explanations. So, I don't think cpr-reboot deserves to be condemned to experimental limbo. - Steve