Re: [PATCH v7 04/15] KVM: vmx/pmu: Emulate MSR_ARCH_LBR_DEPTH for guest Arch LBR

[Yang Weijiang <weijiang.yang@intel.com>]

On Tue, Aug 10, 2021 at 03:54:09PM +0800, Like Xu wrote:
> On 10/8/2021 3:38 pm, Yang Weijiang wrote:
> >On Mon, Aug 09, 2021 at 09:16:47PM +0800, Like Xu wrote:
> >>On 6/8/2021 3:42 pm, Yang Weijiang wrote:
> >>>From: Like Xu <like.xu@linux.intel.com>
> >>
> >>...
> >>
> >>>
> >>>The number of Arch LBR entries available is determined by the value
> >>>in host MSR_ARCH_LBR_DEPTH.DEPTH. The supported LBR depth values are
> >>>enumerated in CPUID.(EAX=01CH, ECX=0):EAX[7:0]. For each bit "n" set
> >>>in this field, the MSR_ARCH_LBR_DEPTH.DEPTH value of "8*(n+1)" is
> >>>supported.
> >>>
> >>>On a guest write to MSR_ARCH_LBR_DEPTH, all LBR entries are reset to 0.
> >>>KVM writes guest requested value to the native ARCH_LBR_DEPTH MSR
> >>>(this is safe because the two values will be the same) when the Arch LBR
> >>>records MSRs are pass-through to the guest.
> >>>
> >>>Signed-off-by: Like Xu <like.xu@linux.intel.com>
> >>>Signed-off-by: Yang Weijiang <weijiang.yang@intel.com>
> >>>---
> >>>  arch/x86/kvm/vmx/pmu_intel.c | 35 ++++++++++++++++++++++++++++++++++-
> >>>  1 file changed, 34 insertions(+), 1 deletion(-)
> >>>
> >>>diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c
> >>>index 9efc1a6b8693..a4ef5bbce186 100644
> >>>--- a/arch/x86/kvm/vmx/pmu_intel.c
> >>>+++ b/arch/x86/kvm/vmx/pmu_intel.c
> >>>@@ -211,7 +211,7 @@ static bool intel_pmu_is_valid_lbr_msr(struct kvm_vcpu *vcpu, u32 index)
> >>>  static bool intel_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
> >>>  {
> >>>  	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >>>-	int ret;
> >>>+	int ret = 0;
> >>>  	switch (msr) {
> >>>  	case MSR_CORE_PERF_FIXED_CTR_CTRL:
> >>>@@ -220,6 +220,10 @@ static bool intel_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
> >>>  	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
> >>>  		ret = pmu->version > 1;
> >>>  		break;
> >>>+	case MSR_ARCH_LBR_DEPTH:
> >>>+		if (kvm_cpu_cap_has(X86_FEATURE_ARCH_LBR))
> >>>+			ret = guest_cpuid_has(vcpu, X86_FEATURE_ARCH_LBR);
> >>>+		break;
> >>>  	default:
> >>>  		ret = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0) ||
> >>>  			get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0) ||
> >>>@@ -348,10 +352,28 @@ static bool intel_pmu_handle_lbr_msrs_access(struct kvm_vcpu *vcpu,
> >>>  	return true;
> >>>  }
> >>>+/*
> >>>+ * Check if the requested depth value the same as that of host.
> >>>+ * When guest/host depth are different, the handling would be tricky,
> >>>+ * so now only max depth is supported for both host and guest.
> >>>+ */
> >>>+static bool arch_lbr_depth_is_valid(struct kvm_vcpu *vcpu, u64 depth)
> >>>+{
> >>>+	unsigned int eax, ebx, ecx, edx;
> >>>+
> >>>+	if (!kvm_cpu_cap_has(X86_FEATURE_ARCH_LBR))
> >>>+		return false;
> >>>+
> >>>+	cpuid_count(0x1c, 0, &eax, &ebx, &ecx, &edx);
> >>
> >>I really don't understand why the sanity check of the
> >>guest lbr depth needs to read the host's cpuid entry and it's pretty slow.
> >>
> >This is to address a concern from Jim:
> >"Does this imply that, when restoring a vCPU, KVM_SET_CPUID2 must be called before
> >KVM_SET_MSRS, so that arch_lbr_depth_is_valid() knows what to do? Is this documented
> >anywhere?"
> 
> What will KVM do if the #GP behaviour of msr does not match the CPUID it is set to?
> 
> For user space host_initiated path, we may check it with the host one but
> for the guest emulation, it should rely on guest cpuid, just like what we do
> in the intel_is_valid_msr().
I think the original consideration is to avoid a mis-match of host/guest depth
makes the host Arch LBR totally broken since a mis-match will clear all LBR MSRs.
But I added a check in kvm_vcpu_after_set_cpuid() when user-space KVM_SET_CPUID2 to
make sure user-space won't set unsupported depth. So maybe I can make
the check as simple as checking a fixed value.
> 
> >anyway, setting depth MSR shouldn't be hot path.
> 
> Not at all, it will be used to reset LBR entries
> and it's as frequent as task switching.

Yes, I saw this, if xsaves is not supported, writing depth msr is used as a short-cut.
> 
> >>KVM has reported the maximum host LBR depth as the only supported value.
> >>
> >>>+
> >>>+	return (depth == fls(eax & 0xff) * 8);
> >>>+}
> >>>+
> >>>  static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >>>  {
> >>>  	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >>>  	struct kvm_pmc *pmc;
> >>>+	struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
> >>>  	u32 msr = msr_info->index;
> >>>  	switch (msr) {
> >>>@@ -367,6 +389,9 @@ static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >>>  	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
> >>>  		msr_info->data = pmu->global_ovf_ctrl;
> >>>  		return 0;
> >>>+	case MSR_ARCH_LBR_DEPTH:
> >>>+		msr_info->data = lbr_desc->records.nr;
> >>>+		return 0;
> >>>  	default:
> >>>  		if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0)) ||
> >>>  		    (pmc = get_gp_pmc(pmu, msr, MSR_IA32_PMC0))) {
> >>>@@ -393,6 +418,7 @@ static int intel_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >>>  {
> >>>  	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >>>  	struct kvm_pmc *pmc;
> >>>+	struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
> >>>  	u32 msr = msr_info->index;
> >>>  	u64 data = msr_info->data;
> >>>@@ -427,6 +453,13 @@ static int intel_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >>>  			return 0;
> >>>  		}
> >>>  		break;
> >>>+	case MSR_ARCH_LBR_DEPTH:
> >>>+		if (!arch_lbr_depth_is_valid(vcpu, data))
> >>>+			return 1;
> >>>+		lbr_desc->records.nr = data;
> >>>+		if (!msr_info->host_initiated)
> >>>+			wrmsrl(MSR_ARCH_LBR_DEPTH, lbr_desc->records.nr);
> >>
> >>Resetting the host msr here is dangerous,
> >>what if the guest LBR event doesn't exist or isn't scheduled on?
> >Hmm, should be vmcs_write to the DEPTH field, thanks for pointing this
> >out!
> 
> Seriously?
My gosh! we don't have the field :-/ Then more sanity checks are
required.
> 
> >>
> >>>+		return 0;
> >>>  	default:
> >>>  		if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0)) ||
> >>>  		    (pmc = get_gp_pmc(pmu, msr, MSR_IA32_PMC0))) {
> >>>