[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <d8b0cea2-9e0b-1622-a253-a17cf642efc7@redhat.com>
Date: Thu, 17 Aug 2017 16:29:38 +0200
From: Paolo Bonzini <pbonzini@...hat.com>
To: Yu Zhang <yu.c.zhang@...ux.intel.com>, kvm@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, rkrcmar@...hat.com,
tglx@...utronix.de, mingo@...hat.com, hpa@...or.com,
xiaoguangrong@...cent.com, joro@...tes.org
Subject: Re: [PATCH v2 1/5] KVM: x86: Add return value to kvm_cpuid().
On 17/08/2017 15:20, Yu Zhang wrote:
>>
>
> OK. And to return 0 for eax/ebx/ecx/edx if check_cpuid_limit() is also
> to be omitted,
> I'd better refactor this patch and move the "out:" before the if
> statement. :-)
>
> best = check_cpuid_limit(vcpu, function, index);
> }
>
> +out:
> if (best) {
> *eax = best->eax;
> *ebx = best->ebx;
> @@ -887,7 +888,6 @@ bool kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax, u32
> *ebx,
> } else
> *eax = *ebx = *ecx = *edx = 0;
>
> -out:
> trace_kvm_cpuid(function, *eax, *ebx, *ecx, *edx, entry_found);
> return entry_found;
> }
>
> And for all get_cpuid() callers which is testing the existence of a
> feature, we do not need to
> check the return value, just checking the flag in the register should be
> fine, correct?
Yes, correct!
Paolo
Powered by blists - more mailing lists