| Message ID | ca5d0399cdbbaa6c7c6528ad85b3560cec0f0752.1708933498.git.isaku.yamahata@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v19,001/130] x86/virt/tdx: Rename _offset to _member for TD_SYSINFO_MAP() macro | expand |
On 2/26/2024 4:26 PM, isaku.yamahata@intel.com wrote: > From: Isaku Yamahata <isaku.yamahata@intel.com> > > Add a flag, KVM_DEBUGREG_AUTO_SWITCHED_GUEST, to skip saving/restoring DRs > irrespective of any other flags. Here "irrespective of any other flags" sounds like other flags will be ignored if KVM_DEBUGREG_AUTO_SWITCHED_GUEST is set. But the code below doesn't align with it. > TDX-SEAM unconditionally saves and > restores guest DRs and reset to architectural INIT state on TD exit. > So, KVM needs to save host DRs before TD enter without restoring guest DRs > and restore host DRs after TD exit. > > Opportunistically convert the KVM_DEBUGREG_* definitions to use BIT(). > > Reported-by: Xiaoyao Li <xiaoyao.li@intel.com> > Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> > Co-developed-by: Chao Gao <chao.gao@intel.com> > Signed-off-by: Chao Gao <chao.gao@intel.com> > Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com> > --- > arch/x86/include/asm/kvm_host.h | 10 ++++++++-- > arch/x86/kvm/vmx/tdx.c | 1 + > arch/x86/kvm/x86.c | 11 ++++++++--- > 3 files changed, 17 insertions(+), 5 deletions(-) > > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h > index 3ab85c3d86ee..a9df898c6fbd 100644 > --- a/arch/x86/include/asm/kvm_host.h > +++ b/arch/x86/include/asm/kvm_host.h > @@ -610,8 +610,14 @@ struct kvm_pmu { > struct kvm_pmu_ops; > > enum { > - KVM_DEBUGREG_BP_ENABLED = 1, > - KVM_DEBUGREG_WONT_EXIT = 2, > + KVM_DEBUGREG_BP_ENABLED = BIT(0), > + KVM_DEBUGREG_WONT_EXIT = BIT(1), > + /* > + * Guest debug registers (DR0-3 and DR6) are saved/restored by hardware > + * on exit from or enter to guest. KVM needn't switch them. Because DR7 > + * is cleared on exit from guest, DR7 need to be saved/restored. > + */ > + KVM_DEBUGREG_AUTO_SWITCH = BIT(2), > }; > > struct kvm_mtrr_range { > diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c > index 7aa9188f384d..ab7403a19c5d 100644 > --- a/arch/x86/kvm/vmx/tdx.c > +++ b/arch/x86/kvm/vmx/tdx.c > @@ -586,6 +586,7 @@ int tdx_vcpu_create(struct kvm_vcpu *vcpu) > > vcpu->arch.efer = EFER_SCE | EFER_LME | EFER_LMA | EFER_NX; > > + vcpu->arch.switch_db_regs = KVM_DEBUGREG_AUTO_SWITCH; > vcpu->arch.cr0_guest_owned_bits = -1ul; > vcpu->arch.cr4_guest_owned_bits = -1ul; > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index 1b189e86a1f1..fb7597c22f31 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -11013,7 +11013,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > if (vcpu->arch.guest_fpu.xfd_err) > wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); > > - if (unlikely(vcpu->arch.switch_db_regs)) { > + if (unlikely(vcpu->arch.switch_db_regs & ~KVM_DEBUGREG_AUTO_SWITCH)) { > set_debugreg(0, 7); > set_debugreg(vcpu->arch.eff_db[0], 0); > set_debugreg(vcpu->arch.eff_db[1], 1); > @@ -11059,6 +11059,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > */ > if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) { > WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP); > + WARN_ON(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH); > static_call(kvm_x86_sync_dirty_debug_regs)(vcpu); > kvm_update_dr0123(vcpu); > kvm_update_dr7(vcpu); > @@ -11071,8 +11072,12 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > * care about the messed up debug address registers. But if > * we have some of them active, restore the old state. > */ > - if (hw_breakpoint_active()) > - hw_breakpoint_restore(); > + if (hw_breakpoint_active()) { > + if (!(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH)) > + hw_breakpoint_restore(); > + else > + set_debugreg(__this_cpu_read(cpu_dr7), 7); According to TDX module 1.5 ABI spec: DR0-3, DR6 and DR7 are set to their architectural INIT value, why is only DR7 restored? I found a discussion about it in https://lore.kernel.org/kvm/Yk4iChT2sbtzSb8h@google.com/ Sean mentioned: " The TDX module context switches the guest _and_ host debug registers. It restores the host DRs because it needs to write _something_ to hide guest state, so it might as well restore the host values. The above was an optmization to avoid rewriting all debug registers. " I have question on "so it might as well restore the host values". Was it a guess/assumption or it was the fact? > + } > > vcpu->arch.last_vmentry_cpu = vcpu->cpu; > vcpu->arch.last_guest_tsc = kvm_read_l1_tsc(vcpu, rdtsc());
On Sun, Apr 07, 2024 at 06:52:44PM +0800, Binbin Wu <binbin.wu@linux.intel.com> wrote: > > > On 2/26/2024 4:26 PM, isaku.yamahata@intel.com wrote: > > From: Isaku Yamahata <isaku.yamahata@intel.com> > > > > Add a flag, KVM_DEBUGREG_AUTO_SWITCHED_GUEST, to skip saving/restoring DRs > > irrespective of any other flags. > > Here "irrespective of any other flags" sounds like other flags will be > ignored if KVM_DEBUGREG_AUTO_SWITCHED_GUEST is set. > But the code below doesn't align with it. Sure, let's update the commit message. > > TDX-SEAM unconditionally saves and > > restores guest DRs and reset to architectural INIT state on TD exit. > > So, KVM needs to save host DRs before TD enter without restoring guest DRs > > and restore host DRs after TD exit. > > > > Opportunistically convert the KVM_DEBUGREG_* definitions to use BIT(). > > > > Reported-by: Xiaoyao Li <xiaoyao.li@intel.com> > > Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> > > Co-developed-by: Chao Gao <chao.gao@intel.com> > > Signed-off-by: Chao Gao <chao.gao@intel.com> > > Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com> > > --- > > arch/x86/include/asm/kvm_host.h | 10 ++++++++-- > > arch/x86/kvm/vmx/tdx.c | 1 + > > arch/x86/kvm/x86.c | 11 ++++++++--- > > 3 files changed, 17 insertions(+), 5 deletions(-) > > > > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h > > index 3ab85c3d86ee..a9df898c6fbd 100644 > > --- a/arch/x86/include/asm/kvm_host.h > > +++ b/arch/x86/include/asm/kvm_host.h > > @@ -610,8 +610,14 @@ struct kvm_pmu { > > struct kvm_pmu_ops; > > enum { > > - KVM_DEBUGREG_BP_ENABLED = 1, > > - KVM_DEBUGREG_WONT_EXIT = 2, > > + KVM_DEBUGREG_BP_ENABLED = BIT(0), > > + KVM_DEBUGREG_WONT_EXIT = BIT(1), > > + /* > > + * Guest debug registers (DR0-3 and DR6) are saved/restored by hardware > > + * on exit from or enter to guest. KVM needn't switch them. Because DR7 > > + * is cleared on exit from guest, DR7 need to be saved/restored. > > + */ > > + KVM_DEBUGREG_AUTO_SWITCH = BIT(2), > > }; > > struct kvm_mtrr_range { > > diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c > > index 7aa9188f384d..ab7403a19c5d 100644 > > --- a/arch/x86/kvm/vmx/tdx.c > > +++ b/arch/x86/kvm/vmx/tdx.c > > @@ -586,6 +586,7 @@ int tdx_vcpu_create(struct kvm_vcpu *vcpu) > > vcpu->arch.efer = EFER_SCE | EFER_LME | EFER_LMA | EFER_NX; > > + vcpu->arch.switch_db_regs = KVM_DEBUGREG_AUTO_SWITCH; > > vcpu->arch.cr0_guest_owned_bits = -1ul; > > vcpu->arch.cr4_guest_owned_bits = -1ul; > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > > index 1b189e86a1f1..fb7597c22f31 100644 > > --- a/arch/x86/kvm/x86.c > > +++ b/arch/x86/kvm/x86.c > > @@ -11013,7 +11013,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > > if (vcpu->arch.guest_fpu.xfd_err) > > wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); > > - if (unlikely(vcpu->arch.switch_db_regs)) { > > + if (unlikely(vcpu->arch.switch_db_regs & ~KVM_DEBUGREG_AUTO_SWITCH)) { > > set_debugreg(0, 7); > > set_debugreg(vcpu->arch.eff_db[0], 0); > > set_debugreg(vcpu->arch.eff_db[1], 1); > > @@ -11059,6 +11059,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > > */ > > if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) { > > WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP); > > + WARN_ON(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH); > > static_call(kvm_x86_sync_dirty_debug_regs)(vcpu); > > kvm_update_dr0123(vcpu); > > kvm_update_dr7(vcpu); > > @@ -11071,8 +11072,12 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > > * care about the messed up debug address registers. But if > > * we have some of them active, restore the old state. > > */ > > - if (hw_breakpoint_active()) > > - hw_breakpoint_restore(); > > + if (hw_breakpoint_active()) { > > + if (!(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH)) > > + hw_breakpoint_restore(); > > + else > > + set_debugreg(__this_cpu_read(cpu_dr7), 7); > > According to TDX module 1.5 ABI spec: > DR0-3, DR6 and DR7 are set to their architectural INIT value, why is only > DR7 restored? This hunk should be dropped. Thank you for finding this. I checked the base SPEC, the ABI spec, and the TDX module code. It seems the documentation bug of the TDX module 1.5 base architecture specification. The TDX module code: - restores guest DR<N> on TD Entry to guest. - saves guest DR<N> on TD Exit from guest TD - initializes DR<N> on TD Exit to host VMM TDX module 1.5 base architecture specification: 15.1.2.1 Context Switch By design, the Intel TDX module context-switches all debug/tracing state that the guest TD is allowed to use. DR0-3, DR6 and IA32_DS_AREA MSR are context-switched in TDH.VP.ENTER and TD exit flows RFLAGS, IA32_DEBUGCTL MSR and DR7 are saved and cleared on VM exits from the guest TD and restored on VM entry to the guest TD. TDX module 1.5 ABI specification: 5.3.65. TDH.VP.ENTER Leaf CPU State Preservation Following a Successful TD Entry and a TD Exit Following a successful TD entry and a TD exit, some CPU state is modified: Registers DR0, DR1, DR2, DR3, DR6 and DR7 are set to their architectural INIT value.
On 2/26/2024 4:26 PM, isaku.yamahata@intel.com wrote: > From: Isaku Yamahata <isaku.yamahata@intel.com> > > Add a flag, KVM_DEBUGREG_AUTO_SWITCHED_GUEST, to skip saving/restoring DRs > irrespective of any other flags. TDX-SEAM unconditionally saves and > restores guest DRs and reset to architectural INIT state on TD exit. > So, KVM needs to save host DRs before TD enter without restoring guest DRs > and restore host DRs after TD exit. > > Opportunistically convert the KVM_DEBUGREG_* definitions to use BIT(). > > Reported-by: Xiaoyao Li <xiaoyao.li@intel.com> > Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> > Co-developed-by: Chao Gao <chao.gao@intel.com> > Signed-off-by: Chao Gao <chao.gao@intel.com> > Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com> > --- > arch/x86/include/asm/kvm_host.h | 10 ++++++++-- > arch/x86/kvm/vmx/tdx.c | 1 + > arch/x86/kvm/x86.c | 11 ++++++++--- > 3 files changed, 17 insertions(+), 5 deletions(-) > > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h > index 3ab85c3d86ee..a9df898c6fbd 100644 > --- a/arch/x86/include/asm/kvm_host.h > +++ b/arch/x86/include/asm/kvm_host.h > @@ -610,8 +610,14 @@ struct kvm_pmu { > struct kvm_pmu_ops; > > enum { > - KVM_DEBUGREG_BP_ENABLED = 1, > - KVM_DEBUGREG_WONT_EXIT = 2, > + KVM_DEBUGREG_BP_ENABLED = BIT(0), > + KVM_DEBUGREG_WONT_EXIT = BIT(1), > + /* > + * Guest debug registers (DR0-3 and DR6) are saved/restored by hardware > + * on exit from or enter to guest. KVM needn't switch them. Because DR7 > + * is cleared on exit from guest, DR7 need to be saved/restored. > + */ > + KVM_DEBUGREG_AUTO_SWITCH = BIT(2), > }; > > struct kvm_mtrr_range { > diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c > index 7aa9188f384d..ab7403a19c5d 100644 > --- a/arch/x86/kvm/vmx/tdx.c > +++ b/arch/x86/kvm/vmx/tdx.c > @@ -586,6 +586,7 @@ int tdx_vcpu_create(struct kvm_vcpu *vcpu) > > vcpu->arch.efer = EFER_SCE | EFER_LME | EFER_LMA | EFER_NX; > > + vcpu->arch.switch_db_regs = KVM_DEBUGREG_AUTO_SWITCH; > vcpu->arch.cr0_guest_owned_bits = -1ul; > vcpu->arch.cr4_guest_owned_bits = -1ul; > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index 1b189e86a1f1..fb7597c22f31 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -11013,7 +11013,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > if (vcpu->arch.guest_fpu.xfd_err) > wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); > > - if (unlikely(vcpu->arch.switch_db_regs)) { > + if (unlikely(vcpu->arch.switch_db_regs & ~KVM_DEBUGREG_AUTO_SWITCH)) { As pointed by Paolo in https://lore.kernel.org/lkml/ea136ac6-53cf-cdc5-a741-acfb437819b1@redhat.com/ KVM_DEBUGREG_BP_ENABLED could be set in vcpu->arch.switch_db_regs, by userspace kvm_vcpu_ioctl_x86_set_debugregs() --> kvm_update_dr7() So it should be fixed as: - if (unlikely(vcpu->arch.switch_db_regs)) { + if (unlikely(vcpu->arch.switch_db_regs && + !(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH))) { > set_debugreg(0, 7); > set_debugreg(vcpu->arch.eff_db[0], 0); > set_debugreg(vcpu->arch.eff_db[1], 1); > @@ -11059,6 +11059,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > */ > if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) { > WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP); > + WARN_ON(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH); > static_call(kvm_x86_sync_dirty_debug_regs)(vcpu); > kvm_update_dr0123(vcpu); > kvm_update_dr7(vcpu); > @@ -11071,8 +11072,12 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > * care about the messed up debug address registers. But if > * we have some of them active, restore the old state. > */ > - if (hw_breakpoint_active()) > - hw_breakpoint_restore(); > + if (hw_breakpoint_active()) { > + if (!(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH)) > + hw_breakpoint_restore(); > + else > + set_debugreg(__this_cpu_read(cpu_dr7), 7); > + } > > vcpu->arch.last_vmentry_cpu = vcpu->cpu; > vcpu->arch.last_guest_tsc = kvm_read_l1_tsc(vcpu, rdtsc());
On Mon, May 06, 2024 at 11:30:52AM +0800, Binbin Wu <binbin.wu@linux.intel.com> wrote: > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > > index 1b189e86a1f1..fb7597c22f31 100644 > > --- a/arch/x86/kvm/x86.c > > +++ b/arch/x86/kvm/x86.c > > @@ -11013,7 +11013,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) > > if (vcpu->arch.guest_fpu.xfd_err) > > wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); > > - if (unlikely(vcpu->arch.switch_db_regs)) { > > + if (unlikely(vcpu->arch.switch_db_regs & ~KVM_DEBUGREG_AUTO_SWITCH)) { > > As pointed by Paolo in > https://lore.kernel.org/lkml/ea136ac6-53cf-cdc5-a741-acfb437819b1@redhat.com/ > KVM_DEBUGREG_BP_ENABLED could be set in vcpu->arch.switch_db_regs, by > userspace > kvm_vcpu_ioctl_x86_set_debugregs() --> kvm_update_dr7() > > So it should be fixed as: > > - if (unlikely(vcpu->arch.switch_db_regs)) { > + if (unlikely(vcpu->arch.switch_db_regs && > + !(vcpu->arch.switch_db_regs & > KVM_DEBUGREG_AUTO_SWITCH))) { Yes, that's true. Thanks for catching this.
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h index 3ab85c3d86ee..a9df898c6fbd 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -610,8 +610,14 @@ struct kvm_pmu { struct kvm_pmu_ops; enum { - KVM_DEBUGREG_BP_ENABLED = 1, - KVM_DEBUGREG_WONT_EXIT = 2, + KVM_DEBUGREG_BP_ENABLED = BIT(0), + KVM_DEBUGREG_WONT_EXIT = BIT(1), + /* + * Guest debug registers (DR0-3 and DR6) are saved/restored by hardware + * on exit from or enter to guest. KVM needn't switch them. Because DR7 + * is cleared on exit from guest, DR7 need to be saved/restored. + */ + KVM_DEBUGREG_AUTO_SWITCH = BIT(2), }; struct kvm_mtrr_range { diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c index 7aa9188f384d..ab7403a19c5d 100644 --- a/arch/x86/kvm/vmx/tdx.c +++ b/arch/x86/kvm/vmx/tdx.c @@ -586,6 +586,7 @@ int tdx_vcpu_create(struct kvm_vcpu *vcpu) vcpu->arch.efer = EFER_SCE | EFER_LME | EFER_LMA | EFER_NX; + vcpu->arch.switch_db_regs = KVM_DEBUGREG_AUTO_SWITCH; vcpu->arch.cr0_guest_owned_bits = -1ul; vcpu->arch.cr4_guest_owned_bits = -1ul; diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 1b189e86a1f1..fb7597c22f31 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -11013,7 +11013,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) if (vcpu->arch.guest_fpu.xfd_err) wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); - if (unlikely(vcpu->arch.switch_db_regs)) { + if (unlikely(vcpu->arch.switch_db_regs & ~KVM_DEBUGREG_AUTO_SWITCH)) { set_debugreg(0, 7); set_debugreg(vcpu->arch.eff_db[0], 0); set_debugreg(vcpu->arch.eff_db[1], 1); @@ -11059,6 +11059,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) */ if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) { WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP); + WARN_ON(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH); static_call(kvm_x86_sync_dirty_debug_regs)(vcpu); kvm_update_dr0123(vcpu); kvm_update_dr7(vcpu); @@ -11071,8 +11072,12 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) * care about the messed up debug address registers. But if * we have some of them active, restore the old state. */ - if (hw_breakpoint_active()) - hw_breakpoint_restore(); + if (hw_breakpoint_active()) { + if (!(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH)) + hw_breakpoint_restore(); + else + set_debugreg(__this_cpu_read(cpu_dr7), 7); + } vcpu->arch.last_vmentry_cpu = vcpu->cpu; vcpu->arch.last_guest_tsc = kvm_read_l1_tsc(vcpu, rdtsc());