From mboxrd@z Thu Jan  1 00:00:00 1970
From: Avi Kivity <avi@redhat.com>
Subject: Re: [PATCH v3] KVM: x86: XSAVE/XRSTOR live migration support
Date: Sun, 13 Jun 2010 11:26:18 +0300
Message-ID: <4C14962A.9030608@redhat.com>
References: <1276230971-5990-1-git-send-email-sheng@linux.intel.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: Marcelo Tosatti <mtosatti@redhat.com>, kvm@vger.kernel.org
To: Sheng Yang <sheng@linux.intel.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:64622 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1753084Ab0FMI4d (ORCPT <rfc822;kvm@vger.kernel.org>);
	Sun, 13 Jun 2010 04:56:33 -0400
In-Reply-To: <1276230971-5990-1-git-send-email-sheng@linux.intel.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On 06/11/2010 07:36 AM, Sheng Yang wrote:
> This patch enable save/restore of xsave state.
>
> +static int kvm_vcpu_ioctl_x86_set_xsave(struct kvm_vcpu *vcpu,
> +					struct kvm_xsave *guest_xsave)
> +{
> +	u64 xstate_bv =
> +		*(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)];
> +	int size;
> +
> +	if (cpu_has_xsave) {
> +		if (xstate_bv&  XSTATE_YMM)
> +			size = XSAVE_YMM_OFFSET + XSAVE_YMM_SIZE;
> +		else
> +			size = XSAVE_HDR_OFFSET + XSAVE_HDR_SIZE;
> +		memcpy(&vcpu->arch.guest_fpu.state->xsave,
> +			guest_xsave->region, size);
>    

This allows userspace to overflow host memory by specifying XSTATE_YMM 
on a host that doesn't support it.

Better to just use the host's size of the structure.

> +	} else {
> +		if (xstate_bv&  ~XSTATE_FPSSE)
> +			return -EINVAL;
> +		size = sizeof(struct i387_fxsave_struct);
> +		memcpy(&vcpu->arch.guest_fpu.state->fxsave,
> +			guest_xsave->region, size);
> +	}
> +	return 0;
> +}
> +
>
> +
> +static int kvm_vcpu_ioctl_x86_set_xcrs(struct kvm_vcpu *vcpu,
> +				       struct kvm_xcrs *guest_xcrs)
> +{
> +	int i, r = 0;
> +
> +	if (!cpu_has_xsave)
> +		return -EINVAL;
>    

Too strict?

> +
> +	if (guest_xcrs->nr_xcrs>  KVM_MAX_XCRS)
> +		return -EFAULT;
>    

EFAULT is for faults during access to userspace.  EINVAL or E2BIG.

Need to ensure flags is 0 for forward compatibility.

> +
> +	for (i = 0; i<  guest_xcrs->nr_xcrs; i++)
> +		/* Only support XCR0 currently */
> +		if (guest_xcrs->xcrs[0].xcr == XCR_XFEATURE_ENABLED_MASK) {
> +			r = __kvm_set_xcr(vcpu, XCR_XFEATURE_ENABLED_MASK,
> +				guest_xcrs->xcrs[0].value);
> +			break;
> +		}
> +	if (r)
> +		r = -EFAULT;
>    

EINVAL

> +	return r;
> +}
> +
>    

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