On Mon, Feb 21, 2022, Chenyi Qiang wrote:
> PKS MSR passes through guest directly. Configure the MSR to match the
> L0/L1 settings so that nested VM runs PKS properly.
>
> Signed-off-by: Chenyi Qiang <chenyi.qiang@xxxxxxxxx>
> ---
> arch/x86/kvm/vmx/nested.c | 38 ++++++++++++++++++++++++++++++++++++--
> arch/x86/kvm/vmx/vmcs12.c | 2 ++
> arch/x86/kvm/vmx/vmcs12.h | 4 ++++
> arch/x86/kvm/vmx/vmx.c | 1 +
> arch/x86/kvm/vmx/vmx.h | 2 ++
> 5 files changed, 45 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index f235f77cbc03..c42a1df385ef 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -252,6 +252,10 @@ static void vmx_sync_vmcs_host_state(struct vcpu_vmx *vmx,
> dest->ds_sel = src->ds_sel;
> dest->es_sel = src->es_sel;
> #endif
> + if (unlikely(src->pkrs != dest->pkrs)) {
> + vmcs_write64(HOST_IA32_PKRS, src->pkrs);
> + dest->pkrs = src->pkrs;
> + }
It's worth adding a helper for this, a la vmx_set_host_fs_gs(), though this one
can probably be an inline in vmx.h. E.g. to yield
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index bfa37c7665a5..906a2913a886 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -252,10 +252,7 @@ static void vmx_sync_vmcs_host_state(struct vcpu_vmx *vmx,
dest->ds_sel = src->ds_sel;
dest->es_sel = src->es_sel;
#endif
- if (unlikely(src->pkrs != dest->pkrs)) {
- vmcs_write64(HOST_IA32_PKRS, src->pkrs);
- dest->pkrs = src->pkrs;
- }
+ vmx_set_host_pkrs(dest, src->pkrs);
}
static void vmx_switch_vmcs(struct kvm_vcpu *vcpu, struct loaded_vmcs *vmcs)
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 35fee600fae7..b6b5f1a46544 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -1157,10 +1157,7 @@ void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
*/
if (vm_exit_controls_get(vmx) & VM_EXIT_LOAD_IA32_PKRS) {
host_pkrs = get_current_pkrs();
- if (unlikely(host_pkrs != host_state->pkrs)) {
- vmcs_write64(HOST_IA32_PKRS, host_pkrs);
- host_state->pkrs = host_pkrs;
- }
+ vmx_set_host_pkrs(host_state, host_pkrs);
}
#ifdef CONFIG_X86_64
> }
>
> static void vmx_switch_vmcs(struct kvm_vcpu *vcpu, struct loaded_vmcs *vmcs)
> @@ -685,6 +689,9 @@ static inline bool nested_vmx_prepare_msr_bitmap(struct kvm_vcpu *vcpu,
> nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
> MSR_IA32_PRED_CMD, MSR_TYPE_W);
>
> + nested_vmx_set_intercept_for_msr(vmx, msr_bitmap_l1, msr_bitmap_l0,
> + MSR_IA32_PKRS, MSR_TYPE_RW);
> +
> kvm_vcpu_unmap(vcpu, &vmx->nested.msr_bitmap_map, false);
>
> vmx->nested.force_msr_bitmap_recalc = false;
> @@ -2433,6 +2440,10 @@ static void prepare_vmcs02_rare(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
> if (kvm_mpx_supported() && vmx->nested.nested_run_pending &&
> (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS))
> vmcs_write64(GUEST_BNDCFGS, vmcs12->guest_bndcfgs);
> +
> + if (vmx->nested.nested_run_pending &&
> + (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PKRS))
> + vmcs_write64(GUEST_IA32_PKRS, vmcs12->guest_ia32_pkrs);
> }
>
> if (nested_cpu_has_xsaves(vmcs12))
> @@ -2521,6 +2532,11 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
> if (kvm_mpx_supported() && (!vmx->nested.nested_run_pending ||
> !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS)))
> vmcs_write64(GUEST_BNDCFGS, vmx->nested.vmcs01_guest_bndcfgs);
> + if (kvm_cpu_cap_has(X86_FEATURE_PKS) &&
> + (!vmx->nested.nested_run_pending ||
> + !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PKRS)))
> + vmcs_write64(GUEST_IA32_PKRS, vmx->nested.vmcs01_guest_pkrs);
> +
> vmx_set_rflags(vcpu, vmcs12->guest_rflags);
>
> /* EXCEPTION_BITMAP and CR0_GUEST_HOST_MASK should basically be the
> @@ -2897,6 +2913,10 @@ static int nested_vmx_check_host_state(struct kvm_vcpu *vcpu,
> vmcs12->host_ia32_perf_global_ctrl)))
> return -EINVAL;
>
> + if ((vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PKRS) &&
> + CC(!kvm_pkrs_valid(vmcs12->host_ia32_pkrs)))
Please align the indentation:
if ((vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PKRS) &&
CC(!kvm_pkrs_valid(vmcs12->host_ia32_pkrs)))
return -EINVAL;
> + return -EINVAL;
> +
> #ifdef CONFIG_X86_64
> ia32e = !!(vmcs12->vm_exit_controls & VM_EXIT_HOST_ADDR_SPACE_SIZE);
> #else
> @@ -3049,6 +3069,10 @@ static int nested_vmx_check_guest_state(struct kvm_vcpu *vcpu,
> if (nested_check_guest_non_reg_state(vmcs12))
> return -EINVAL;
>
> + if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PKRS) &&
> + CC(!kvm_pkrs_valid(vmcs12->guest_ia32_pkrs)))
> + return -EINVAL;
> +
> return 0;
> }
>
> @@ -3377,6 +3401,9 @@ enum nvmx_vmentry_status nested_vmx_enter_non_root_mode(struct kvm_vcpu *vcpu,
> if (kvm_mpx_supported() &&
> !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS))
> vmx->nested.vmcs01_guest_bndcfgs = vmcs_read64(GUEST_BNDCFGS);
> + if (kvm_cpu_cap_has(X86_FEATURE_PKS) &&
> + !(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PKRS))
This needs read the current PKRS if from_vmentry == false, e.g.
if (kvm_cpu_cap_has(X86_FEATURE_PKS) &&
(!from_vmentry ||
!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PKRS)))
because in the migration case, if nested state is set after MSR state, the value
needs to come from the current MSR value, which was propagated to vmc02 (which
this calls vmcs01, but whatever).
Note, I'm pretty sure the GUEST_BNDCFGS code is broken, surprise surprise.
> + vmx->nested.vmcs01_guest_pkrs = vmcs_read64(GUEST_IA32_PKRS);
>
> /*
> * Overwrite vmcs01.GUEST_CR3 with L1's CR3 if EPT is disabled *and*
> @@ -4022,6 +4049,7 @@ static bool is_vmcs12_ext_field(unsigned long field)
> case GUEST_IDTR_BASE:
> case GUEST_PENDING_DBG_EXCEPTIONS:
> case GUEST_BNDCFGS:
> + case GUEST_IA32_PKRS:
> return true;
> default:
> break;
> @@ -4073,6 +4101,8 @@ static void sync_vmcs02_to_vmcs12_rare(struct kvm_vcpu *vcpu,
> vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS);
> if (kvm_mpx_supported())
> vmcs12->guest_bndcfgs = vmcs_read64(GUEST_BNDCFGS);
> + if (guest_cpuid_has(vcpu, X86_FEATURE_PKS))
This needs to check vmx->nested.msrs.entry_ctls_* (I can never remember if it's
the high or low part...). The SDM states PKRS is saved "if the processor supports
the 1-setting of the 'load PKRS' VM-entry control", which is different than PKRS
being supported in CPUID. Also, guest CPUID is userspace controlled, e.g. userspace
could induce a failed VMREAD by giving a garbage CPUID model, where vmx->nested.msrs
can only be restricted by userspace, i.e. is trusted.
Happyily, checking vmx->nested.msrs is also a performance win, as guest_cpuid_has()
can require walking a large array.
