From: Jim Mattson <jmattson@google.com>
To: Like Xu <like.xu.linux@gmail.com>
Cc: Sean Christopherson <seanjc@google.com>,
Paolo Bonzini <pbonzini@redhat.com>,
Sandipan Das <sandipan.das@amd.com>,
kvm@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 3/4] KVM: x86/svm/pmu: Add AMD PerfMonV2 support
Date: Tue, 6 Sep 2022 13:19:45 -0700 [thread overview]
Message-ID: <CALMp9eSQ1QkmECM4at9XDPUew0h2nxG5=YUSN=aWnQpZkXy2dw@mail.gmail.com> (raw)
In-Reply-To: <c9b3d50e-ec3d-3fa3-2706-5672100ffe09@gmail.com>
On Tue, Sep 6, 2022 at 5:45 AM Like Xu <like.xu.linux@gmail.com> wrote:
>
> On 6/9/2022 2:00 am, Jim Mattson wrote:
> > On Mon, Sep 5, 2022 at 5:44 AM Like Xu <like.xu.linux@gmail.com> wrote:
> >>
> >> From: Like Xu <likexu@tencent.com>
> >>
> >> If AMD Performance Monitoring Version 2 (PerfMonV2) is detected
> >> by the guest, it can use a new scheme to manage the Core PMCs using
> >> the new global control and status registers.
> >>
> >> In addition to benefiting from the PerfMonV2 functionality in the same
> >> way as the host (higher precision), the guest also can reduce the number
> >> of vm-exits by lowering the total number of MSRs accesses.
> >>
> >> In terms of implementation details, amd_is_valid_msr() is resurrected
> >> since three newly added MSRs could not be mapped to one vPMC.
> >> The possibility of emulating PerfMonV2 on the mainframe has also
> >> been eliminated for reasons of precision.
> >>
> >> Co-developed-by: Sandipan Das <sandipan.das@amd.com>
> >> Signed-off-by: Sandipan Das <sandipan.das@amd.com>
> >> Signed-off-by: Like Xu <likexu@tencent.com>
> >> ---
> >> arch/x86/kvm/pmu.c | 6 +++++
> >> arch/x86/kvm/svm/pmu.c | 50 +++++++++++++++++++++++++++++++++---------
> >> arch/x86/kvm/x86.c | 11 ++++++++++
> >> 3 files changed, 57 insertions(+), 10 deletions(-)
> >>
> >> diff --git a/arch/x86/kvm/pmu.c b/arch/x86/kvm/pmu.c
> >> index 7002e1b74108..56b4f898a246 100644
> >> --- a/arch/x86/kvm/pmu.c
> >> +++ b/arch/x86/kvm/pmu.c
> >> @@ -455,12 +455,15 @@ int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >>
> >> switch (msr) {
> >> case MSR_CORE_PERF_GLOBAL_STATUS:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS:
> >> msr_info->data = pmu->global_status;
> >> return 0;
> >> case MSR_CORE_PERF_GLOBAL_CTRL:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_CTL:
> >> msr_info->data = pmu->global_ctrl;
> >> return 0;
> >> case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS_CLR:
> >> msr_info->data = 0;
> >> return 0;
> >> default:
> >> @@ -479,12 +482,14 @@ int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >>
> >> switch (msr) {
> >> case MSR_CORE_PERF_GLOBAL_STATUS:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS:
> >> if (msr_info->host_initiated) {
> >> pmu->global_status = data;
> >> return 0;
> >> }
> >> break; /* RO MSR */
> >> case MSR_CORE_PERF_GLOBAL_CTRL:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_CTL:
> >> if (pmu->global_ctrl == data)
> >> return 0;
> >> if (kvm_valid_perf_global_ctrl(pmu, data)) {
> >> @@ -495,6 +500,7 @@ int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >> }
> >> break;
> >> case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS_CLR:
> >> if (!(data & pmu->global_ovf_ctrl_mask)) {
> >> if (!msr_info->host_initiated)
> >> pmu->global_status &= ~data;
> >> diff --git a/arch/x86/kvm/svm/pmu.c b/arch/x86/kvm/svm/pmu.c
> >> index 3a20972e9f1a..4c7d408e3caa 100644
> >> --- a/arch/x86/kvm/svm/pmu.c
> >> +++ b/arch/x86/kvm/svm/pmu.c
> >> @@ -92,12 +92,6 @@ static struct kvm_pmc *amd_rdpmc_ecx_to_pmc(struct kvm_vcpu *vcpu,
> >> return amd_pmc_idx_to_pmc(vcpu_to_pmu(vcpu), idx & ~(3u << 30));
> >> }
> >>
> >> -static bool amd_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
> >> -{
> >> - /* All MSRs refer to exactly one PMC, so msr_idx_to_pmc is enough. */
> >> - return false;
> >> -}
> >> -
> >> static struct kvm_pmc *amd_msr_idx_to_pmc(struct kvm_vcpu *vcpu, u32 msr)
> >> {
> >> struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >> @@ -109,6 +103,29 @@ static struct kvm_pmc *amd_msr_idx_to_pmc(struct kvm_vcpu *vcpu, u32 msr)
> >> return pmc;
> >> }
> >>
> >> +static bool amd_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
> >> +{
> >> + struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >> +
> >> + switch (msr) {
> >> + case MSR_K7_EVNTSEL0 ... MSR_K7_PERFCTR3:
> >> + return pmu->version > 0;
> >> + case MSR_F15H_PERF_CTL0 ... MSR_F15H_PERF_CTR5:
> >> + return guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE);
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_CTL:
> >> + case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS_CLR:
> >> + return pmu->version > 1;
> >> + default:
> >> + if (msr > MSR_F15H_PERF_CTR5 &&
> >> + msr < MSR_F15H_PERF_CTL0 + 2 * KVM_AMD_PMC_MAX_GENERIC)
> >> + return pmu->version > 1;
> >
> > Should this be bounded by guest CPUID.80000022H:EBX[NumCorePmc]
> > (unless host-initiated)?
>
> Indeed, how about:
>
> default:
> if (msr > MSR_F15H_PERF_CTR5 &&
> msr < MSR_F15H_PERF_CTL0 + 2 * pmu->nr_arch_gp_counters)
> return pmu->version > 1;
>
> and for host-initiated:
>
> #define MSR_F15H_PERF_MSR_MAX \
> (MSR_F15H_PERF_CTR0 + 2 * (KVM_AMD_PMC_MAX_GENERIC - 1))
I think there may be an off-by-one error here.
>
> kvm_{set|get}_msr_common()
> case MSR_F15H_PERF_CTL0 ... MSR_F15H_PERF_MSR_MAX:
> if (kvm_pmu_is_valid_msr(vcpu, msr))
> return kvm_pmu_set_msr(vcpu, msr_info);
> ?
>
> >
> >> + break;
> >> + }
> >> +
> >> + return amd_msr_idx_to_pmc(vcpu, msr);
> >> +}
> >> +
> >> static int amd_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >> {
> >> struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >> @@ -162,20 +179,31 @@ static int amd_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >> static void amd_pmu_refresh(struct kvm_vcpu *vcpu)
> >> {
> >> struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
> >> + struct kvm_cpuid_entry2 *entry;
> >> + union cpuid_0x80000022_ebx ebx;
> >>
> >> - if (guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE))
> >> + pmu->version = 1;
> >> + entry = kvm_find_cpuid_entry_index(vcpu, 0x80000022, 0);
> >> + if (kvm_pmu_cap.version > 1 && entry && (entry->eax & BIT(0))) {
> >> + pmu->version = 2;
> >> + ebx.full = entry->ebx;
> >> + pmu->nr_arch_gp_counters = min3((unsigned int)ebx.split.num_core_pmc,
> >> + (unsigned int)kvm_pmu_cap.num_counters_gp,
> >> + (unsigned int)KVM_AMD_PMC_MAX_GENERIC);
> >> + pmu->global_ctrl_mask = ~((1ull << pmu->nr_arch_gp_counters) - 1);
> >> + pmu->global_ovf_ctrl_mask = pmu->global_ctrl_mask;
> >> + } else if (guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE)) {
> >> pmu->nr_arch_gp_counters = AMD64_NUM_COUNTERS_CORE;
> >
> > The logic above doesn't seem quite right, since guest_cpuid_has(vcpu,
> > X86_FEATURE_PERFCTR_CORE) promises 6 PMCs, regardless of what
> > CPUID.80000022 says.
>
> I would have expected the appearance of CPUID.80000022 to override PERFCTR_CORE,
> now I don't think it's a good idea as you do, so how about:
>
> amd_pmu_refresh():
>
> bool perfctr_core = guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE);
>
> pmu->version = 1;
> if (kvm_pmu_cap.version > 1)
> entry = kvm_find_cpuid_entry_index(vcpu, 0x80000022, 0);
>
> if (!perfctr_core)
> pmu->nr_arch_gp_counters = AMD64_NUM_COUNTERS;
> if (entry && (entry->eax & BIT(0))) {
> pmu->version = 2;
> ebx.full = entry->ebx;
> pmu->nr_arch_gp_counters = min3((unsigned int)ebx.split.num_core_pmc,
> (unsigned int)kvm_pmu_cap.num_counters_gp,
> (unsigned int)KVM_AMD_PMC_MAX_GENERIC);
> }
> /* PERFCTR_CORE promises 6 PMCs, regardless of CPUID.80000022 */
> if (perfctr_core) {
> pmu->nr_arch_gp_counters = max(pmu->nr_arch_gp_counters,
> AMD64_NUM_COUNTERS_CORE);
> }
Even if X86_FEATURE_PERFCTR_CORE is clear, all AMD CPUs promise 4 PMCs.
>
> if (pmu->version > 1) {
> pmu->global_ctrl_mask = ~((1ull << pmu->nr_arch_gp_counters) - 1);
> pmu->global_ovf_ctrl_mask = pmu->global_ctrl_mask;
> }
>
> ?
>
>
next prev parent reply other threads:[~2022-09-06 20:21 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-05 12:39 [PATCH 0/4] KVM: x86/svm/pmu: Add AMD Guest PerfMonV2 support Like Xu
2022-09-05 12:39 ` [PATCH 1/4] KVM: x86/svm/pmu: Limit the maximum number of supported GP counters Like Xu
2022-09-05 17:26 ` Jim Mattson
2022-09-06 12:38 ` Like Xu
2022-09-05 12:39 ` [PATCH 2/4] KVM: x86/pmu: Make part of the Intel v2 PMU MSRs handling x86 generic Like Xu
2022-09-05 12:39 ` [PATCH 3/4] KVM: x86/svm/pmu: Add AMD PerfMonV2 support Like Xu
2022-09-05 18:00 ` Jim Mattson
2022-09-06 12:45 ` Like Xu
2022-09-06 20:19 ` Jim Mattson [this message]
2022-09-07 3:50 ` Like Xu
2022-09-07 4:15 ` Jim Mattson
2022-09-05 12:39 ` [PATCH 4/4] KVM: x86/cpuid: Add AMD CPUID ExtPerfMonAndDbg leaf 0x80000022 Like Xu
2022-09-05 17:36 ` Jim Mattson
2022-09-06 12:53 ` Like Xu
2022-09-06 20:08 ` Jim Mattson
2022-09-07 3:59 ` Like Xu
2022-09-07 4:11 ` Jim Mattson
2022-09-07 5:52 ` Sandipan Das
2022-09-07 6:39 ` Like Xu
2022-09-08 6:00 ` Sandipan Das
2022-09-08 23:14 ` Jim Mattson
2022-09-05 12:39 ` [kvm-unit-tests PATCH 1/2] x86/pmu: Update rdpmc testcase to cover #GP and emulation path Like Xu
2022-10-05 21:36 ` Sean Christopherson
2022-10-19 8:50 ` Like Xu
2022-09-05 12:39 ` [kvm-unit-tests PATCH 2/2] x86/pmu: Add AMD Guest PerfMonV2 testcases Like Xu
2022-10-05 22:08 ` Sean Christopherson
2022-10-19 9:40 ` Like Xu
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='CALMp9eSQ1QkmECM4at9XDPUew0h2nxG5=YUSN=aWnQpZkXy2dw@mail.gmail.com' \
--to=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=sandipan.das@amd.com \
--cc=seanjc@google.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).