From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Cooper Subject: Re: [PATCH] VMX: Eliminate cr3 store/load vmexit when UG enabled Date: Wed, 23 Oct 2013 15:40:39 +0100 Message-ID: <5267DFE7.5000802@citrix.com> References: <1382513945-9479-1-git-send-email-yang.z.zhang@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1382513945-9479-1-git-send-email-yang.z.zhang@intel.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Yang Zhang Cc: xen-devel@lists.xensource.com, eddie.dong@intel.com, JBeulich@suse.com List-Id: xen-devel@lists.xenproject.org On 23/10/13 08:39, Yang Zhang wrote: > From: Yang Zhang > > With the feature of unrestricted guest, there should no vmexit > be triggered when guest accesses the cr3 in non-paging mode. > > Signed-off-by: Yang Zhang You English here confused me for a bit. I presume you mean "Xen should not cause vmexits for cr3 accesses in unrestricted guests", whereas the current meaning implies that hardware wont generate a vmexit for cr3 accesses for unrestricted guests (which is not correct according to the SDM). > --- > xen/arch/x86/hvm/vmx/vmx.c | 6 +++--- > 1 files changed, 3 insertions(+), 3 deletions(-) > > diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c > index 9ca8632..b9b5e74 100644 > --- a/xen/arch/x86/hvm/vmx/vmx.c > +++ b/xen/arch/x86/hvm/vmx/vmx.c > @@ -1086,7 +1086,7 @@ static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr) > uint32_t cr3_ctls = (CPU_BASED_CR3_LOAD_EXITING | > CPU_BASED_CR3_STORE_EXITING); > v->arch.hvm_vmx.exec_control &= ~cr3_ctls; > - if ( !hvm_paging_enabled(v) ) > + if ( !hvm_paging_enabled(v) && !vmx_unrestricted_guest(v) ) >>From my (brief, so please correct me if I am wrong) reading of the spec, guest_CR0.PG is enforced if SECONDARY_EXEC_UNRESTRICTED_GUEST is clear. Therefore, !hvm_paging_enable() implies SECONDARY_EXEC_UNRESTRICTED_GUEST is set (or we will incur a vmentry failure) > v->arch.hvm_vmx.exec_control |= cr3_ctls; > > /* Trap CR3 updates if CR3 memory events are enabled. */ > @@ -1156,7 +1156,7 @@ static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr) > case 3: > if ( paging_mode_hap(v->domain) ) > { > - if ( !hvm_paging_enabled(v) ) > + if ( !hvm_paging_enabled(v) && !vmx_unrestricted_guest(v) ) > v->arch.hvm_vcpu.hw_cr[3] = > v->domain->arch.hvm_domain.params[HVM_PARAM_IDENT_PT]; > vmx_load_pdptrs(v); Why does unrestricted mode affect which pagetables we use when moving back into Xen? > @@ -2408,7 +2408,7 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs) > > hvm_invalidate_regs_fields(regs); > > - if ( paging_mode_hap(v->domain) && hvm_paging_enabled(v) ) > + if ( paging_mode_hap(v->domain) ) > { > __vmread(GUEST_CR3, &v->arch.hvm_vcpu.hw_cr[3]); > v->arch.hvm_vcpu.guest_cr[3] = v->arch.hvm_vcpu.hw_cr[3]; ~Andrew