Re: [PATCH 03/13] accel: Elide kvm_update_guest_debug by checking kvm_supports_guest_debug

[Richard Henderson]

On 2/9/22 02:22, Philippe Mathieu-Daudé wrote:
> -int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap)
> -{
> -    return -ENOSYS;
> -}
> -
>   int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
>                             target_ulong len, int type)
>   {
> diff --git a/cpu.c b/cpu.c
> index 97d42b6b2a..834e2b4cdb 100644
> --- a/cpu.c
> +++ b/cpu.c
> @@ -350,7 +350,7 @@ void cpu_single_step(CPUState *cpu, int enabled)
>   {
>       if (cpu->singlestep_enabled != enabled) {
>           cpu->singlestep_enabled = enabled;
> -        if (kvm_enabled()) {
> +        if (kvm_enabled() && kvm_supports_guest_debug()) {
>               kvm_update_guest_debug(cpu, 0);

Why is this second hunk required?

I would think the stub would be removable on its own, since it is protected by 
kvm_enabled, which will expand to false exactly like kvm_supports_guest_debug expands to 
false.

As an aside, I wonder about the non-checking of the update return value...


r~