On 13/06/2016 12:06, Oscar Fernando Garcia wrote: > Good Morning, > > I have a simple question. In KVM code: /arch/x86/kvm/x86.c when the > function vcpu_enter_guest is called, the run function ( > kvm_x86_ops->run(vcpu)) is called with preemption and irqs disabled. > My question is, running in non-root mode with preemption and irqs > disabled affects the latency of other processes or VPCUs running in the > same CPU? Even though kvm_x86_ops->run itself is called with interrupts disabled, interrupts are enabled while the guest runs. Instead of being serviced immediately by the processor, however, they cause a vmexit. The interrupt routine is then called during kvm_x86_ops->handle_external_intr, and preemption is enabled again soon after that in vcpu_enter_guest. The extra latency introduced by KVM is very low, around 1000 clock cycles or less than a microsecond. Paolo -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
