Hi Prasad,
On 1/2/24 23:53, Prasad Pandit wrote:
> From: Prasad Pandit <pjp@xxxxxxxxxxxxxxxxx>
>
> kvm_vcpu_ioctl_x86_set_vcpu_events() routine makes 'KVM_REQ_NMI'
> request for a vcpu even when its 'events->nmi.pending' is zero.
> Ex:
> qemu_thread_start
> kvm_vcpu_thread_fn
> qemu_wait_io_event
> qemu_wait_io_event_common
> process_queued_cpu_work
> do_kvm_cpu_synchronize_post_init/_reset
> kvm_arch_put_registers
> kvm_put_vcpu_events (cpu, level=[2|3])
>
> This leads vCPU threads in QEMU to constantly acquire & release the
> global mutex lock, delaying the guest boot due to lock contention.
Would you mind sharing where and how the lock contention is at QEMU space? That
is, how the QEMU mutex lock is impacted by KVM KVM_REQ_NMI?
Or you meant line 3031 at QEMU side?
2858 int kvm_cpu_exec(CPUState *cpu)
2859 {
2860 struct kvm_run *run = cpu->kvm_run;
2861 int ret, run_ret;
... ...
3023 default:
3024 DPRINTF("kvm_arch_handle_exit\n");
3025 ret = kvm_arch_handle_exit(cpu, run);
3026 break;
3027 }
3028 } while (ret == 0);
3029
3030 cpu_exec_end(cpu);
3031 qemu_mutex_lock_iothread();
3032
3033 if (ret < 0) {
3034 cpu_dump_state(cpu, stderr, CPU_DUMP_CODE);
3035 vm_stop(RUN_STATE_INTERNAL_ERROR);
3036 }
3037
3038 qatomic_set(&cpu->exit_request, 0);
3039 return ret;
3040 }
Thank you very much!
Dongli Zhang
> Add check to make KVM_REQ_NMI request only if vcpu has NMI pending.
>
> Fixes: bdedff263132 ("KVM: x86: Route pending NMIs from userspace through process_nmi()")
> Signed-off-by: Prasad Pandit <pjp@xxxxxxxxxxxxxxxxx>
> ---
> arch/x86/kvm/x86.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 1a3aaa7dafae..468870450b8b 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -5405,7 +5405,8 @@ static int kvm_vcpu_ioctl_x86_set_vcpu_events(struct kvm_vcpu *vcpu,
> if (events->flags & KVM_VCPUEVENT_VALID_NMI_PENDING) {
> vcpu->arch.nmi_pending = 0;
> atomic_set(&vcpu->arch.nmi_queued, events->nmi.pending);
> - kvm_make_request(KVM_REQ_NMI, vcpu);
> + if (events->nmi.pending)
> + kvm_make_request(KVM_REQ_NMI, vcpu);
> }
> static_call(kvm_x86_set_nmi_mask)(vcpu, events->nmi.masked);
>
> --
> 2.43.0
>
>
