linux/arch/x86/kvm/vmx/pmu_intel.c

402 lines
10 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* KVM PMU support for Intel CPUs
*
* Copyright 2011 Red Hat, Inc. and/or its affiliates.
*
* Authors:
* Avi Kivity <avi@redhat.com>
* Gleb Natapov <gleb@redhat.com>
*/
#include <linux/types.h>
#include <linux/kvm_host.h>
#include <linux/perf_event.h>
#include <asm/perf_event.h>
#include "x86.h"
#include "cpuid.h"
#include "lapic.h"
#include "nested.h"
#include "pmu.h"
static struct kvm_event_hw_type_mapping intel_arch_events[] = {
/* Index must match CPUID 0x0A.EBX bit vector */
[0] = { 0x3c, 0x00, PERF_COUNT_HW_CPU_CYCLES },
[1] = { 0xc0, 0x00, PERF_COUNT_HW_INSTRUCTIONS },
[2] = { 0x3c, 0x01, PERF_COUNT_HW_BUS_CYCLES },
[3] = { 0x2e, 0x4f, PERF_COUNT_HW_CACHE_REFERENCES },
[4] = { 0x2e, 0x41, PERF_COUNT_HW_CACHE_MISSES },
[5] = { 0xc4, 0x00, PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
[6] = { 0xc5, 0x00, PERF_COUNT_HW_BRANCH_MISSES },
[7] = { 0x00, 0x30, PERF_COUNT_HW_REF_CPU_CYCLES },
};
/* mapping between fixed pmc index and intel_arch_events array */
static int fixed_pmc_events[] = {1, 0, 7};
static void reprogram_fixed_counters(struct kvm_pmu *pmu, u64 data)
{
int i;
for (i = 0; i < pmu->nr_arch_fixed_counters; i++) {
u8 new_ctrl = fixed_ctrl_field(data, i);
u8 old_ctrl = fixed_ctrl_field(pmu->fixed_ctr_ctrl, i);
struct kvm_pmc *pmc;
pmc = get_fixed_pmc(pmu, MSR_CORE_PERF_FIXED_CTR0 + i);
if (old_ctrl == new_ctrl)
continue;
__set_bit(INTEL_PMC_IDX_FIXED + i, pmu->pmc_in_use);
reprogram_fixed_counter(pmc, new_ctrl, i);
}
pmu->fixed_ctr_ctrl = data;
}
/* function is called when global control register has been updated. */
static void global_ctrl_changed(struct kvm_pmu *pmu, u64 data)
{
int bit;
u64 diff = pmu->global_ctrl ^ data;
pmu->global_ctrl = data;
for_each_set_bit(bit, (unsigned long *)&diff, X86_PMC_IDX_MAX)
reprogram_counter(pmu, bit);
}
static unsigned intel_find_arch_event(struct kvm_pmu *pmu,
u8 event_select,
u8 unit_mask)
{
int i;
for (i = 0; i < ARRAY_SIZE(intel_arch_events); i++)
if (intel_arch_events[i].eventsel == event_select
&& intel_arch_events[i].unit_mask == unit_mask
&& (pmu->available_event_types & (1 << i)))
break;
if (i == ARRAY_SIZE(intel_arch_events))
return PERF_COUNT_HW_MAX;
return intel_arch_events[i].event_type;
}
static unsigned intel_find_fixed_event(int idx)
{
if (idx >= ARRAY_SIZE(fixed_pmc_events))
return PERF_COUNT_HW_MAX;
return intel_arch_events[fixed_pmc_events[idx]].event_type;
}
/* check if a PMC is enabled by comparing it with globl_ctrl bits. */
static bool intel_pmc_is_enabled(struct kvm_pmc *pmc)
{
struct kvm_pmu *pmu = pmc_to_pmu(pmc);
return test_bit(pmc->idx, (unsigned long *)&pmu->global_ctrl);
}
static struct kvm_pmc *intel_pmc_idx_to_pmc(struct kvm_pmu *pmu, int pmc_idx)
{
if (pmc_idx < INTEL_PMC_IDX_FIXED)
return get_gp_pmc(pmu, MSR_P6_EVNTSEL0 + pmc_idx,
MSR_P6_EVNTSEL0);
else {
u32 idx = pmc_idx - INTEL_PMC_IDX_FIXED;
return get_fixed_pmc(pmu, idx + MSR_CORE_PERF_FIXED_CTR0);
}
}
/* returns 0 if idx's corresponding MSR exists; otherwise returns 1. */
static int intel_is_valid_rdpmc_ecx(struct kvm_vcpu *vcpu, unsigned int idx)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
bool fixed = idx & (1u << 30);
idx &= ~(3u << 30);
return (!fixed && idx >= pmu->nr_arch_gp_counters) ||
(fixed && idx >= pmu->nr_arch_fixed_counters);
}
static struct kvm_pmc *intel_rdpmc_ecx_to_pmc(struct kvm_vcpu *vcpu,
unsigned int idx, u64 *mask)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
bool fixed = idx & (1u << 30);
struct kvm_pmc *counters;
idx &= ~(3u << 30);
if (!fixed && idx >= pmu->nr_arch_gp_counters)
return NULL;
if (fixed && idx >= pmu->nr_arch_fixed_counters)
return NULL;
counters = fixed ? pmu->fixed_counters : pmu->gp_counters;
*mask &= pmu->counter_bitmask[fixed ? KVM_PMC_FIXED : KVM_PMC_GP];
return &counters[idx];
}
static bool intel_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
int ret;
switch (msr) {
case MSR_CORE_PERF_FIXED_CTR_CTRL:
case MSR_CORE_PERF_GLOBAL_STATUS:
case MSR_CORE_PERF_GLOBAL_CTRL:
case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
ret = pmu->version > 1;
break;
default:
ret = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0) ||
get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0) ||
get_fixed_pmc(pmu, msr);
break;
}
return ret;
}
static struct kvm_pmc *intel_msr_idx_to_pmc(struct kvm_vcpu *vcpu, u32 msr)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
struct kvm_pmc *pmc;
pmc = get_fixed_pmc(pmu, msr);
pmc = pmc ? pmc : get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0);
pmc = pmc ? pmc : get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0);
return pmc;
}
static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *data)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
struct kvm_pmc *pmc;
switch (msr) {
case MSR_CORE_PERF_FIXED_CTR_CTRL:
*data = pmu->fixed_ctr_ctrl;
return 0;
case MSR_CORE_PERF_GLOBAL_STATUS:
*data = pmu->global_status;
return 0;
case MSR_CORE_PERF_GLOBAL_CTRL:
*data = pmu->global_ctrl;
return 0;
case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
*data = pmu->global_ovf_ctrl;
return 0;
default:
if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0))) {
u64 val = pmc_read_counter(pmc);
*data = val & pmu->counter_bitmask[KVM_PMC_GP];
return 0;
} else if ((pmc = get_fixed_pmc(pmu, msr))) {
u64 val = pmc_read_counter(pmc);
*data = val & pmu->counter_bitmask[KVM_PMC_FIXED];
return 0;
} else if ((pmc = get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0))) {
*data = pmc->eventsel;
return 0;
}
}
return 1;
}
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;
u32 msr = msr_info->index;
u64 data = msr_info->data;
switch (msr) {
case MSR_CORE_PERF_FIXED_CTR_CTRL:
if (pmu->fixed_ctr_ctrl == data)
return 0;
if (!(data & 0xfffffffffffff444ull)) {
reprogram_fixed_counters(pmu, data);
return 0;
}
break;
case MSR_CORE_PERF_GLOBAL_STATUS:
if (msr_info->host_initiated) {
pmu->global_status = data;
return 0;
}
break; /* RO MSR */
case MSR_CORE_PERF_GLOBAL_CTRL:
if (pmu->global_ctrl == data)
return 0;
if (kvm_valid_perf_global_ctrl(pmu, data)) {
global_ctrl_changed(pmu, data);
return 0;
}
break;
case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
if (!(data & pmu->global_ovf_ctrl_mask)) {
if (!msr_info->host_initiated)
pmu->global_status &= ~data;
pmu->global_ovf_ctrl = data;
return 0;
}
break;
default:
if ((pmc = get_gp_pmc(pmu, msr, MSR_IA32_PERFCTR0))) {
if (msr_info->host_initiated)
pmc->counter = data;
else
pmc->counter = (s32)data;
return 0;
} else if ((pmc = get_fixed_pmc(pmu, msr))) {
pmc->counter = data;
return 0;
} else if ((pmc = get_gp_pmc(pmu, msr, MSR_P6_EVNTSEL0))) {
if (data == pmc->eventsel)
return 0;
if (!(data & pmu->reserved_bits)) {
reprogram_gp_counter(pmc, data);
return 0;
}
}
}
return 1;
}
static void intel_pmu_refresh(struct kvm_vcpu *vcpu)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
struct x86_pmu_capability x86_pmu;
struct kvm_cpuid_entry2 *entry;
union cpuid10_eax eax;
union cpuid10_edx edx;
pmu->nr_arch_gp_counters = 0;
pmu->nr_arch_fixed_counters = 0;
pmu->counter_bitmask[KVM_PMC_GP] = 0;
pmu->counter_bitmask[KVM_PMC_FIXED] = 0;
pmu->version = 0;
pmu->reserved_bits = 0xffffffff00200000ull;
entry = kvm_find_cpuid_entry(vcpu, 0xa, 0);
if (!entry)
return;
eax.full = entry->eax;
edx.full = entry->edx;
pmu->version = eax.split.version_id;
if (!pmu->version)
return;
perf_get_x86_pmu_capability(&x86_pmu);
pmu->nr_arch_gp_counters = min_t(int, eax.split.num_counters,
x86_pmu.num_counters_gp);
pmu->counter_bitmask[KVM_PMC_GP] = ((u64)1 << eax.split.bit_width) - 1;
pmu->available_event_types = ~entry->ebx &
((1ull << eax.split.mask_length) - 1);
if (pmu->version == 1) {
pmu->nr_arch_fixed_counters = 0;
} else {
pmu->nr_arch_fixed_counters =
min_t(int, edx.split.num_counters_fixed,
x86_pmu.num_counters_fixed);
pmu->counter_bitmask[KVM_PMC_FIXED] =
((u64)1 << edx.split.bit_width_fixed) - 1;
}
pmu->global_ctrl = ((1ull << pmu->nr_arch_gp_counters) - 1) |
(((1ull << pmu->nr_arch_fixed_counters) - 1) << INTEL_PMC_IDX_FIXED);
pmu->global_ctrl_mask = ~pmu->global_ctrl;
pmu->global_ovf_ctrl_mask = pmu->global_ctrl_mask
& ~(MSR_CORE_PERF_GLOBAL_OVF_CTRL_OVF_BUF |
MSR_CORE_PERF_GLOBAL_OVF_CTRL_COND_CHGD);
if (kvm_x86_ops->pt_supported())
pmu->global_ovf_ctrl_mask &=
~MSR_CORE_PERF_GLOBAL_OVF_CTRL_TRACE_TOPA_PMI;
entry = kvm_find_cpuid_entry(vcpu, 7, 0);
if (entry &&
(boot_cpu_has(X86_FEATURE_HLE) || boot_cpu_has(X86_FEATURE_RTM)) &&
(entry->ebx & (X86_FEATURE_HLE|X86_FEATURE_RTM)))
pmu->reserved_bits ^= HSW_IN_TX|HSW_IN_TX_CHECKPOINTED;
bitmap_set(pmu->all_valid_pmc_idx,
0, pmu->nr_arch_gp_counters);
bitmap_set(pmu->all_valid_pmc_idx,
INTEL_PMC_MAX_GENERIC, pmu->nr_arch_fixed_counters);
nested_vmx_pmu_entry_exit_ctls_update(vcpu);
}
static void intel_pmu_init(struct kvm_vcpu *vcpu)
{
int i;
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
for (i = 0; i < INTEL_PMC_MAX_GENERIC; i++) {
pmu->gp_counters[i].type = KVM_PMC_GP;
pmu->gp_counters[i].vcpu = vcpu;
pmu->gp_counters[i].idx = i;
KVM: x86/vPMU: Reuse perf_event to avoid unnecessary pmc_reprogram_counter The perf_event_create_kernel_counter() in the pmc_reprogram_counter() is a heavyweight and high-frequency operation, especially when host disables the watchdog (maximum 21000000 ns) which leads to an unacceptable latency of the guest NMI handler. It limits the use of vPMUs in the guest. When a vPMC is fully enabled, the legacy reprogram_*_counter() would stop and release its existing perf_event (if any) every time EVEN in most cases almost the same requested perf_event will be created and configured again. For each vPMC, if the reuqested config ('u64 eventsel' for gp and 'u8 ctrl' for fixed) is the same as its current config AND a new sample period based on pmc->counter is accepted by host perf interface, the current event could be reused safely as a new created one does. Otherwise, do release the undesirable perf_event and reprogram a new one as usual. It's light-weight to call pmc_pause_counter (disable, read and reset event) and pmc_resume_counter (recalibrate period and re-enable event) as guest expects instead of release-and-create again on any condition. Compared to use the filterable event->attr or hw.config, a new 'u64 current_config' field is added to save the last original programed config for each vPMC. Based on this implementation, the number of calls to pmc_reprogram_counter is reduced by ~82.5% for a gp sampling event and ~99.9% for a fixed event. In the usage of multiplexing perf sampling mode, the average latency of the guest NMI handler is reduced from 104923 ns to 48393 ns (~2.16x speed up). If host disables watchdog, the minimum latecy of guest NMI handler could be speed up at ~3413x (from 20407603 to 5979 ns) and at ~786x in the average. Suggested-by: Kan Liang <kan.liang@linux.intel.com> Signed-off-by: Like Xu <like.xu@linux.intel.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2019-10-27 10:52:42 +00:00
pmu->gp_counters[i].current_config = 0;
}
for (i = 0; i < INTEL_PMC_MAX_FIXED; i++) {
pmu->fixed_counters[i].type = KVM_PMC_FIXED;
pmu->fixed_counters[i].vcpu = vcpu;
pmu->fixed_counters[i].idx = i + INTEL_PMC_IDX_FIXED;
KVM: x86/vPMU: Reuse perf_event to avoid unnecessary pmc_reprogram_counter The perf_event_create_kernel_counter() in the pmc_reprogram_counter() is a heavyweight and high-frequency operation, especially when host disables the watchdog (maximum 21000000 ns) which leads to an unacceptable latency of the guest NMI handler. It limits the use of vPMUs in the guest. When a vPMC is fully enabled, the legacy reprogram_*_counter() would stop and release its existing perf_event (if any) every time EVEN in most cases almost the same requested perf_event will be created and configured again. For each vPMC, if the reuqested config ('u64 eventsel' for gp and 'u8 ctrl' for fixed) is the same as its current config AND a new sample period based on pmc->counter is accepted by host perf interface, the current event could be reused safely as a new created one does. Otherwise, do release the undesirable perf_event and reprogram a new one as usual. It's light-weight to call pmc_pause_counter (disable, read and reset event) and pmc_resume_counter (recalibrate period and re-enable event) as guest expects instead of release-and-create again on any condition. Compared to use the filterable event->attr or hw.config, a new 'u64 current_config' field is added to save the last original programed config for each vPMC. Based on this implementation, the number of calls to pmc_reprogram_counter is reduced by ~82.5% for a gp sampling event and ~99.9% for a fixed event. In the usage of multiplexing perf sampling mode, the average latency of the guest NMI handler is reduced from 104923 ns to 48393 ns (~2.16x speed up). If host disables watchdog, the minimum latecy of guest NMI handler could be speed up at ~3413x (from 20407603 to 5979 ns) and at ~786x in the average. Suggested-by: Kan Liang <kan.liang@linux.intel.com> Signed-off-by: Like Xu <like.xu@linux.intel.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2019-10-27 10:52:42 +00:00
pmu->fixed_counters[i].current_config = 0;
}
}
static void intel_pmu_reset(struct kvm_vcpu *vcpu)
{
struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
struct kvm_pmc *pmc = NULL;
int i;
for (i = 0; i < INTEL_PMC_MAX_GENERIC; i++) {
pmc = &pmu->gp_counters[i];
pmc_stop_counter(pmc);
pmc->counter = pmc->eventsel = 0;
}
for (i = 0; i < INTEL_PMC_MAX_FIXED; i++) {
pmc = &pmu->fixed_counters[i];
pmc_stop_counter(pmc);
pmc->counter = 0;
}
pmu->fixed_ctr_ctrl = pmu->global_ctrl = pmu->global_status =
pmu->global_ovf_ctrl = 0;
}
struct kvm_pmu_ops intel_pmu_ops = {
.find_arch_event = intel_find_arch_event,
.find_fixed_event = intel_find_fixed_event,
.pmc_is_enabled = intel_pmc_is_enabled,
.pmc_idx_to_pmc = intel_pmc_idx_to_pmc,
.rdpmc_ecx_to_pmc = intel_rdpmc_ecx_to_pmc,
.msr_idx_to_pmc = intel_msr_idx_to_pmc,
.is_valid_rdpmc_ecx = intel_is_valid_rdpmc_ecx,
.is_valid_msr = intel_is_valid_msr,
.get_msr = intel_pmu_get_msr,
.set_msr = intel_pmu_set_msr,
.refresh = intel_pmu_refresh,
.init = intel_pmu_init,
.reset = intel_pmu_reset,
};