From: Yang Weijiang <weijiang.yang@intel.com>
To: Like Xu <like.xu.linux@gmail.com>
Cc: Yang Weijiang <weijiang.yang@intel.com>,
pbonzini@redhat.com, jmattson@google.com, seanjc@google.com,
vkuznets@redhat.com, wei.w.wang@intel.com, kvm@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v7 04/15] KVM: vmx/pmu: Emulate MSR_ARCH_LBR_DEPTH for guest Arch LBR
Date: Tue, 10 Aug 2021 17:08:30 +0800 [thread overview]
Message-ID: <20210810090830.GA3120@intel.com> (raw)
In-Reply-To: <0707a301-b6b5-0f54-810f-ef07e4148943@gmail.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))) {
> >>>
next prev parent reply other threads:[~2021-08-10 8:55 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-06 7:42 [PATCH v7 00/15] Introduce Architectural LBR for vPMU Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 01/15] perf/x86/intel: Fix the comment about guest LBR support on KVM Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 02/15] perf/x86/lbr: Simplify the exposure check for the LBR_INFO registers Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 03/15] KVM: x86: Add Arch LBR MSRs to msrs_to_save_all list Yang Weijiang
2021-08-09 13:07 ` Like Xu
2021-08-06 7:42 ` [PATCH v7 04/15] KVM: vmx/pmu: Emulate MSR_ARCH_LBR_DEPTH for guest Arch LBR Yang Weijiang
2021-08-09 13:16 ` Like Xu
2021-08-10 7:38 ` Yang Weijiang
2021-08-10 7:54 ` Like Xu
2021-08-10 9:08 ` Yang Weijiang [this message]
2021-08-06 7:42 ` [PATCH v7 05/15] KVM: vmx/pmu: Emulate MSR_ARCH_LBR_CTL " Yang Weijiang
2021-08-06 16:26 ` kernel test robot
2021-08-09 13:36 ` Like Xu
2021-08-10 8:30 ` Yang Weijiang
2021-08-10 8:37 ` Like Xu
2021-08-06 7:42 ` [PATCH v7 06/15] KVM: x86/pmu: Refactor code to support " Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 07/15] KVM: x86: Refresh CPUID on writes to MSR_IA32_XSS Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 08/15] KVM: x86: Report XSS as an MSR to be saved if there are supported features Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 09/15] KVM: x86: Refine the matching and clearing logic for supported_xss Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 10/15] KVM: x86: Add XSAVE Support for Architectural LBR Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 11/15] KVM: x86/vmx: Check Arch LBR config when return perf capabilities Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 12/15] KVM: nVMX: Add necessary Arch LBR settings for nested VM Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 13/15] KVM: x86/vmx: Clear Arch LBREn bit before inject #DB to guest Yang Weijiang
2021-08-09 5:08 ` Like Xu
2021-08-09 9:02 ` Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 14/15] KVM: x86/vmx: Flip Arch LBREn bit on guest state change Yang Weijiang
2021-08-06 7:42 ` [PATCH v7 15/15] KVM: x86/cpuid: Advise Arch LBR feature in CPUID Yang Weijiang
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210810090830.GA3120@intel.com \
--to=weijiang.yang@intel.com \
--cc=jmattson@google.com \
--cc=kvm@vger.kernel.org \
--cc=like.xu.linux@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=seanjc@google.com \
--cc=vkuznets@redhat.com \
--cc=wei.w.wang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).