Re: [PATCH 2/6] KVM: x86: move all vcpu->arch.pio* setup in emulator_pio_in_out

[Sean Christopherson <seanjc@google.com>]

On Wed, Jun 08, 2022, Paolo Bonzini wrote:
> For now, this is basically an excuse to add back the void* argument to
> the function, while removing some knowledge of vcpu->arch.pio* from
> its callers.  The WARN that vcpu->arch.pio.count is zero is also
> extended to OUT operations.
> 
> We cannot do more as long as we have __emulator_pio_in always followed

Please add parantheses when referencing functions in shortlogs and changelogs,
I find it tremendously helpful.

> by complete_emulator_pio_in, which uses the vcpu->arch.pio* fields.
> But after fixing that, it will be possible to only populate the
> vcpu->arch.pio* fields on userspace exits.

Same nits about about pronouns.  In a similar vein, be explicit about what "more"
mean; I had no idea what "more" meant until the second sentence.  E.g.

  The vcpu->arch.pio* fields still need to be filled even when the PIO is
  handled in-kernel as __emulator_pio_in() is always followed by
  complete_emulator_pio_in().  But after fixing that, it will be possible to
  to only populate the vcpu->arch.pio* fields on userspace exits.

> No functional change intended.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  arch/x86/kvm/trace.h |  2 +-
>  arch/x86/kvm/x86.c   | 18 ++++++++++--------
>  2 files changed, 11 insertions(+), 9 deletions(-)
> 
> diff --git a/arch/x86/kvm/trace.h b/arch/x86/kvm/trace.h
> index fd28dd40b813..2877c0e92823 100644
> --- a/arch/x86/kvm/trace.h
> +++ b/arch/x86/kvm/trace.h
> @@ -154,7 +154,7 @@ TRACE_EVENT(kvm_xen_hypercall,
>  
>  TRACE_EVENT(kvm_pio,
>  	TP_PROTO(unsigned int rw, unsigned int port, unsigned int size,
> -		 unsigned int count, void *data),
> +		 unsigned int count, const void *data),
>  	TP_ARGS(rw, port, size, count, data),
>  
>  	TP_STRUCT__entry(
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 2f9100f2564e..8e1e76d0378b 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -7416,17 +7416,22 @@ static int emulator_cmpxchg_emulated(struct x86_emulate_ctxt *ctxt,
>  }
>  
>  static int emulator_pio_in_out(struct kvm_vcpu *vcpu, int size,
> -			       unsigned short port,
> +			       unsigned short port, void *data,
>  			       unsigned int count, bool in)
>  {
> -	void *data = vcpu->arch.pio_data;
>  	unsigned i;
>  	int r;
>  
> +	WARN_ON_ONCE(vcpu->arch.pio.count);
>  	vcpu->arch.pio.port = port;
>  	vcpu->arch.pio.in = in;
>  	vcpu->arch.pio.count = count;
>  	vcpu->arch.pio.size = size;
> +	if (in)
> +		memset(vcpu->arch.pio_data, 0, size * count);
> +	else
> +		memcpy(vcpu->arch.pio_data, data, size * count);
> +	data = vcpu->arch.pio_data;

Oof, passing NULL for @data and then overwriting it below is gross.  It also makes
@in redundant for this one patch.  Might be worth adding a comment, even though
it's transient?

>  
>  	for (i = 0; i < count; i++) {
>  		if (in)
> @@ -7454,9 +7459,7 @@ static int emulator_pio_in_out(struct kvm_vcpu *vcpu, int size,
>  static int __emulator_pio_in(struct kvm_vcpu *vcpu, int size,
>  			     unsigned short port, unsigned int count)
>  {
> -	WARN_ON(vcpu->arch.pio.count);
> -	memset(vcpu->arch.pio_data, 0, size * count);
> -	return emulator_pio_in_out(vcpu, size, port, count, true);
> +	return emulator_pio_in_out(vcpu, size, port, NULL, count, true);
>  }
>  
>  static void complete_emulator_pio_in(struct kvm_vcpu *vcpu, void *val)
> @@ -7505,9 +7508,8 @@ static int emulator_pio_out(struct kvm_vcpu *vcpu, int size,
>  {
>  	int ret;
>  
> -	memcpy(vcpu->arch.pio_data, val, size * count);
> -	trace_kvm_pio(KVM_PIO_OUT, port, size, count, vcpu->arch.pio_data);
> -	ret = emulator_pio_in_out(vcpu, size, port, count, false);
> +	trace_kvm_pio(KVM_PIO_OUT, port, size, count, val);
> +	ret = emulator_pio_in_out(vcpu, size, port, (void *)val, count, false);
>  	if (ret)
>                  vcpu->arch.pio.count = 0;
>  
> -- 
> 2.31.1
> 
>