From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Hildenbrand Subject: Re: [GIT PULL 7/9] KVM: s390: Backup the guest's machine check info Date: Wed, 28 Jun 2017 21:46:07 +0200 Message-ID: <2659ff50-b264-b0cb-ce61-e604653b6e90@redhat.com> References: <1498671044-81240-1-git-send-email-borntraeger@de.ibm.com> <1498671044-81240-8-git-send-email-borntraeger@de.ibm.com> <5d3bbee4-b3cf-ad98-828c-779624c3a138@redhat.com> <58cf662e-a7c0-2c4f-8d60-402ab9b9edfe@de.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <58cf662e-a7c0-2c4f-8d60-402ab9b9edfe@de.ibm.com> Content-Language: en-US Sender: kvm-owner@vger.kernel.org List-Archive: List-Post: To: Christian Borntraeger , Paolo Bonzini , =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= Cc: KVM , Cornelia Huck , linux-s390 , QingFeng Hao List-ID: > No. Several guest registers are shared with host registers and reloaded lazily. > For example if the floating point registers are invalid, they will be invalid > for the host (but the registers contain the guest values). > From a KVM perspective we can always turn off valid bits in case of doubts > (overindicate errors). So the current variant is certainly not perfect, but it > at least will not claim correct registers when they are incorrect. > We can certainly improve the handling in a future patch. (e.g. turn on validity > for control registers), I will have a look. > As far as I understand, that would then mean, enabling all validity flags for registers that are not shared. Makes sense. -- Thanks, David