Avi Kivity wrote: > Jan Kiszka wrote: >> We already spent to much electrons and brain cycles on possibly "much >> simpler" workarounds. I think injecting and handling a single-step, even >> while there is guest debugging going on or the guest itself single-steps >> or both, will not be more complicated - but "more correct". >> > > I agree. I'm still worried about interactions between the IRET single > stepping code and other things which use the debug registers.
The interaction is inside KVM, so under our control. We may simply save the related states, hook into the guest-debugging parts that evaluate single step exceptions, and handle this case with highest prio, transparently to users of lower prio. In theory - you never really know until you tried to implement it... Jan
signature.asc
Description: OpenPGP digital signature
