[PATCH] RISC-V: KVM: Add kvm-riscv.wfi_trap_policy to control VS-mode WFI trapping

[Anup Patel]

On Thu, Jul 9, 2026 at 5:26 PM Yuhang.chen <yhchen312 at gmail.com> wrote:
>
> Add a kernel command-line option, kvm-riscv.wfi_trap_policy=trap|notrap,
> that controls whether a WFI executed by a VS-mode guest traps into KVM
> (HS-mode) or executes natively.
>
> Measured results (wfi_exit_stat delta / guest wake count over 3 s):
>
>   policy        WFI_EXITS   WAKE_CNT
>   ----------    ---------   --------
>   default       295         294      (== trap, no regression)
>   trap          295         294
>   notrap          0         298
>
> Assisted-by: YuanSheng:deepseek-v4-pro
> Co-developed-by: Quan Zhou <zhouquan at iscas.ac.cn>
> Signed-off-by: Quan Zhou <zhouquan at iscas.ac.cn>
> Signed-off-by: Yuhang.chen <yhchen312 at gmail.com>
> ---
>  .../admin-guide/kernel-parameters.txt         | 14 ++++++
>  arch/riscv/kvm/vcpu.c                         | 44 ++++++++++++++++++-
>  2 files changed, 57 insertions(+), 1 deletion(-)
>
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index b5493a7f8f22..ee6603b30033 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -3254,6 +3254,20 @@ Kernel parameters
>
>                         notrap: clear WFI instruction trap
>
> +       kvm-riscv.wfi_trap_policy=
> +                       [KVM,RISCV] Control when to set the WFI instruction
> +                       trap for KVM VMs. When set, a VS-mode WFI traps into
> +                       KVM and is emulated; when clear, the guest executes
> +                       WFI natively and blocks until a VS-mode interrupt
> +                       (e.g. the sstc timer) is pending.
> +
> +                       trap: set WFI instruction trap (HSTATUS.VTW=1)
> +
> +                       notrap: clear WFI instruction trap (HSTATUS.VTW=0)
> +
> +                       Defaults to trap, preserving the previous
> +                       unconditional behavior.
> +

This would mean that for the "notrap" policy the VCPU would never
sleep and increase host CPU utilization even when VCPU is idle.
Is this expected behaviour ?

Regards,
Anup