On Friday, October 28, 2022 2:03 AM, Sean Christopherson wrote: > But why does it matter what pCPU a vCPU is running on? Wouldn't the > probability of triggering a race between kvm_on_user_return() and > hardware_disable() be _higher_ if there are more pCPUs returning to userspace? I think the point there is that the vcpus and those syscall threads need to be on the same pCPUs. Linux by default has its own load balancing for threads to run on. If the vcpus and syscall threads are scattered on different pCPUs, kvm_on_user_return would less likely to be triggered when the syscall threads return to userspace.
