[PATCH 3/3] arm64: kvm: Fix single step for guest skipped instructions

[Paolo Bonzini]

On 04/10/2017 12:08, Alex Bennée wrote:
> 
> From 2e8fcea695a9eca67fbeb331d3104d1d9e7e337a Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?Alex=20Benn=C3=A9e?= <alex.bennee at linaro.org>
> Date: Wed, 4 Oct 2017 09:49:41 +0000
> Subject: [PATCH] kvm: exit run loop after emulating IO when single stepping
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
> 
> If single-stepping is enabled we should exit the run-loop after
> emulating the access. Otherwise single-stepping across emulated IO
> accesses may skip an instruction.
> 
> This only addresses user-space emulation. Stuff done in kernel-mode
> should be handled there.
> 
> Signed-off-by: Alex Bennée <alex.bennee at linaro.org>
> ---
>  accel/kvm/kvm-all.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> index 90c88b517d..85bcb2b0d4 100644
> --- a/accel/kvm/kvm-all.c
> +++ b/accel/kvm/kvm-all.c
> @@ -1940,7 +1940,7 @@ int kvm_cpu_exec(CPUState *cpu)
>                            run->io.direction,
>                            run->io.size,
>                            run->io.count);
> -            ret = 0;
> +            ret = cpu->singlestep_enabled ? EXCP_DEBUG : 0;
>              break;
>          case KVM_EXIT_MMIO:
>              DPRINTF("handle_mmio\n");
> @@ -1950,7 +1950,7 @@ int kvm_cpu_exec(CPUState *cpu)
>                               run->mmio.data,
>                               run->mmio.len,
>                               run->mmio.is_write);
> -            ret = 0;
> +            ret = cpu->singlestep_enabled ? EXCP_DEBUG : 0;
>              break;
>          case KVM_EXIT_IRQ_WINDOW_OPEN:
>              DPRINTF("irq_window_open\n");

Singlestep mode doesn't make much sense for KVM.  For TCG the purpose is
to build one-instruction translation blocks, but what would it mean for KVM?

Paolo