From mboxrd@z Thu Jan  1 00:00:00 1970
From: Avi Kivity <avi@redhat.com>
Subject: Re: [PATCH 5/5] kvm/svm: copy instruction bytes from VMCB
Date: Tue, 07 Dec 2010 15:33:22 +0200
Message-ID: <4CFE37A2.7010703@redhat.com>
References: <1291719586-22533-1-git-send-email-andre.przywara@amd.com> <1291719586-22533-6-git-send-email-andre.przywara@amd.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: kvm@vger.kernel.org, mtosatti@redhat.com
To: Andre Przywara <andre.przywara@amd.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:32382 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1752806Ab0LGNd1 (ORCPT <rfc822;kvm@vger.kernel.org>);
	Tue, 7 Dec 2010 08:33:27 -0500
In-Reply-To: <1291719586-22533-6-git-send-email-andre.przywara@amd.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On 12/07/2010 12:59 PM, Andre Przywara wrote:
> In case of a nested page fault or an intercepted #PF newer SVM
> implementations provide a copy of the faulting instruction bytes
> in the VMCB.
> Use these bytes to feed the instruction emulator and avoid the costly
> guest instruction fetch in this case.
>
>
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index cfbcbfa..3e3a67e 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -586,6 +586,9 @@ struct kvm_x86_ops {
>   	void (*write_tsc_offset)(struct kvm_vcpu *vcpu, u64 offset);
>
>   	void (*get_exit_info)(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2);
> +
> +	int (*prefetch_instruction)(struct kvm_vcpu *vcpu);
> +
>   	const struct trace_print_flags *exit_reasons_str;
>   };

How about adding a byte array/len parameter to x86_decode_insn()?  It 
could be used to prefill the buffer instead of invoking a callback.

>
> diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
> index 6366735..abff8ff 100644
> --- a/arch/x86/kvm/emulate.c
> +++ b/arch/x86/kvm/emulate.c
> @@ -525,6 +525,7 @@ static int do_insn_fetch(struct x86_emulate_ctxt *ctxt,
>   	/* x86 instructions are limited to 15 bytes. */
>   	if (eip + size - ctxt->eip>  15)
>   		return X86EMUL_UNHANDLEABLE;
> +	kvm_x86_ops->prefetch_instruction(ctxt->vcpu);

Even with the callback, this belongs in x86_decode_insn(), not on every 
fetch.

>   	while (size--) {
>   		rc = do_fetch_insn_byte(ctxt, ops, eip++, dest++);
>   		if (rc != X86EMUL_CONTINUE)
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 3cf2cef..ed94e9a 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -464,6 +464,24 @@ static void skip_emulated_instruction(struct kvm_vcpu *vcpu)
>   	svm_set_interrupt_shadow(vcpu, 0);
>   }
>
> +static int prefetch_instruction(struct kvm_vcpu *vcpu)

svm_prefetch_instruction()

> +{
> +	struct vcpu_svm *svm = to_svm(vcpu);
> +	uint8_t len;
> +	struct fetch_cache *fetch;
> +
> +	len = svm->vmcb->control.insn_len&  0x0F;
> +	if (len == 0)
> +		return 1;
> +
> +	fetch =&svm->vcpu.arch.emulate_ctxt.decode.fetch;
> +	fetch->start = kvm_rip_read(&svm->vcpu);
> +	fetch->end = fetch->start + len;
> +	memcpy(fetch->data, svm->vmcb->control.insn_bytes, len);
> +
> +	return 0;
> +}

svm code shouldn't reach in so deep into the emulator privates.

-- 
error compiling committee.c: too many arguments to function