* [PATCH v7 0/3] KVM: x86: Provide per VM capability for disabling PMU virtualization
@ 2022-02-15 1:48 David Dunn
2022-02-15 1:48 ` [PATCH v7 1/3] " David Dunn
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: David Dunn @ 2022-02-15 1:48 UTC (permalink / raw)
To: pbonzini; +Cc: seanjc, jmattson, like.xu.linux, kvm, David Dunn
Allow usermode to disable PMU virtualization on individual x86 VMs.
When disabled, the PMU is not accessible from the guest.
v6: https://lore.kernel.org/all/20220209172945.1495014-1-daviddunn@google.com/
v6 -> v7
* Incorporating feedback from Sean.
* Rephrased commit messages (borrowed from Sean).
* Added locking around kvm->created_vcpus access.
* Moved KVM_CAP_PMU_VALID_MASK out of user ABI headers.
* Changed comments on vm_create_without_vcpus.
* Directly check results in selftest TEST_ASSERT statements.
v5 -> v6
* resolve minor conflicts that were queued after v5 was reviewed.
v4 -> v5
* Remove automatic CPUID adjustment when PMU disabled.
* Update documentation and changelog to reflect above.
* Update documentation to document arg[0] and return values.
David Dunn (3):
KVM: x86: Provide per VM capability for disabling PMU virtualization
KVM: selftests: Carve out helper to create "default" VM without vCPUs
KVM: selftests: Verify disabling PMU virtualization via
KVM_CAP_CONFIG_PMU
Documentation/virt/kvm/api.rst | 22 ++++++++++++
arch/x86/include/asm/kvm_host.h | 1 +
arch/x86/kvm/svm/pmu.c | 2 +-
arch/x86/kvm/vmx/pmu_intel.c | 2 +-
arch/x86/kvm/x86.c | 17 +++++++++
include/uapi/linux/kvm.h | 3 ++
tools/include/uapi/linux/kvm.h | 3 ++
.../selftests/kvm/include/kvm_util_base.h | 3 ++
tools/testing/selftests/kvm/lib/kvm_util.c | 35 +++++++++++++++----
.../kvm/x86_64/pmu_event_filter_test.c | 33 +++++++++++++++++
10 files changed, 113 insertions(+), 8 deletions(-)
--
2.35.1.265.g69c8d7142f-goog
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v7 1/3] KVM: x86: Provide per VM capability for disabling PMU virtualization
2022-02-15 1:48 [PATCH v7 0/3] KVM: x86: Provide per VM capability for disabling PMU virtualization David Dunn
@ 2022-02-15 1:48 ` David Dunn
2022-02-17 16:22 ` Sean Christopherson
2022-02-15 1:48 ` [PATCH v7 2/3] KVM: selftests: Carve out helper to create "default" VM without vCPUs David Dunn
2022-02-15 1:48 ` [PATCH v7 3/3] KVM: selftests: Verify disabling PMU virtualization via KVM_CAP_CONFIG_PMU David Dunn
2 siblings, 1 reply; 6+ messages in thread
From: David Dunn @ 2022-02-15 1:48 UTC (permalink / raw)
To: pbonzini; +Cc: seanjc, jmattson, like.xu.linux, kvm, David Dunn
Add a new capability, KVM_CAP_PMU_CAPABILITY, that takes a bitmask of
settings/features to allow userspace to configure PMU virtualization on
a per-VM basis. For now, support a single flag, KVM_CAP_PMU_DISABLE,
to allow disabling PMU virtualization for a VM even when KVM is configured
with enable_pmu=true a module level.
To keep KVM simple, disallow changing VM's PMU configuration after vCPUs
have been created.
Signed-off-by: David Dunn <daviddunn@google.com>
---
Documentation/virt/kvm/api.rst | 22 ++++++++++++++++++++++
arch/x86/include/asm/kvm_host.h | 1 +
arch/x86/kvm/svm/pmu.c | 2 +-
arch/x86/kvm/vmx/pmu_intel.c | 2 +-
arch/x86/kvm/x86.c | 17 +++++++++++++++++
include/uapi/linux/kvm.h | 3 +++
tools/include/uapi/linux/kvm.h | 3 +++
7 files changed, 48 insertions(+), 2 deletions(-)
diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
index a4267104db50..df836965b347 100644
--- a/Documentation/virt/kvm/api.rst
+++ b/Documentation/virt/kvm/api.rst
@@ -7561,3 +7561,25 @@ The argument to KVM_ENABLE_CAP is also a bitmask, and must be a subset
of the result of KVM_CHECK_EXTENSION. KVM will forward to userspace
the hypercalls whose corresponding bit is in the argument, and return
ENOSYS for the others.
+
+8.35 KVM_CAP_PMU_CAPABILITY
+---------------------------
+
+:Capability KVM_CAP_PMU_CAPABILITY
+:Architectures: x86
+:Type: vm
+:Parameters: arg[0] is bitmask of PMU virtualization capabilities.
+:Returns 0 on success, -EINVAL when arg[0] contains invalid bits
+
+This capability alters PMU virtualization in KVM.
+
+Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of
+PMU virtualization capabilities that can be adjusted on a VM.
+
+The argument to KVM_ENABLE_CAP is also a bitmask and selects specific
+PMU virtualization capabilities to be applied to the VM. This can
+only be invoked on a VM prior to the creation of VCPUs.
+
+At this time, KVM_CAP_PMU_DISABLE is the only capability. Setting
+this capability will disable PMU virtualization for that VM. Usermode
+should adjust CPUID leaf 0xA to reflect that the PMU is disabled.
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 10815b672a26..61e9050a3488 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1233,6 +1233,7 @@ struct kvm_arch {
hpa_t hv_root_tdp;
spinlock_t hv_root_tdp_lock;
#endif
+ bool enable_pmu;
};
struct kvm_vm_stat {
diff --git a/arch/x86/kvm/svm/pmu.c b/arch/x86/kvm/svm/pmu.c
index 5aa45f13b16d..d4de52409335 100644
--- a/arch/x86/kvm/svm/pmu.c
+++ b/arch/x86/kvm/svm/pmu.c
@@ -101,7 +101,7 @@ static inline struct kvm_pmc *get_gp_pmc_amd(struct kvm_pmu *pmu, u32 msr,
{
struct kvm_vcpu *vcpu = pmu_to_vcpu(pmu);
- if (!enable_pmu)
+ if (!vcpu->kvm->arch.enable_pmu)
return NULL;
switch (msr) {
diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c
index 03fab48b149c..4e5b1eeeb77c 100644
--- a/arch/x86/kvm/vmx/pmu_intel.c
+++ b/arch/x86/kvm/vmx/pmu_intel.c
@@ -487,7 +487,7 @@ static void intel_pmu_refresh(struct kvm_vcpu *vcpu)
pmu->reserved_bits = 0xffffffff00200000ull;
entry = kvm_find_cpuid_entry(vcpu, 0xa, 0);
- if (!entry || !enable_pmu)
+ if (!entry || !vcpu->kvm->arch.enable_pmu)
return;
eax.full = entry->eax;
edx.full = entry->edx;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index eaa3b5b89c5e..0803a1388bbf 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -110,6 +110,8 @@ static u64 __read_mostly cr4_reserved_bits = CR4_RESERVED_BITS;
#define KVM_EXIT_HYPERCALL_VALID_MASK (1 << KVM_HC_MAP_GPA_RANGE)
+#define KVM_CAP_PMU_VALID_MASK KVM_CAP_PMU_DISABLE
+
#define KVM_X2APIC_API_VALID_FLAGS (KVM_X2APIC_API_USE_32BIT_IDS | \
KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK)
@@ -4331,6 +4333,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
if (r < sizeof(struct kvm_xsave))
r = sizeof(struct kvm_xsave);
break;
+ case KVM_CAP_PMU_CAPABILITY:
+ r = enable_pmu ? KVM_CAP_PMU_VALID_MASK : 0;
+ break;
}
default:
break;
@@ -6005,6 +6010,17 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
kvm->arch.exit_on_emulation_error = cap->args[0];
r = 0;
break;
+ case KVM_CAP_PMU_CAPABILITY:
+ r = -EINVAL;
+ if (!enable_pmu || (cap->args[0] & ~KVM_CAP_PMU_VALID_MASK))
+ break;
+ mutex_lock(&kvm->lock);
+ if (!kvm->created_vcpus) {
+ kvm->arch.enable_pmu = !(cap->args[0] & KVM_CAP_PMU_DISABLE);
+ r = 0;
+ }
+ mutex_unlock(&kvm->lock);
+ break;
default:
r = -EINVAL;
break;
@@ -11591,6 +11607,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags);
kvm->arch.guest_can_read_msr_platform_info = true;
+ kvm->arch.enable_pmu = enable_pmu;
#if IS_ENABLED(CONFIG_HYPERV)
spin_lock_init(&kvm->arch.hv_root_tdp_lock);
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index 5191b57e1562..3d6046295755 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -1134,6 +1134,7 @@ struct kvm_ppc_resize_hpt {
#define KVM_CAP_VM_GPA_BITS 207
#define KVM_CAP_XSAVE2 208
#define KVM_CAP_SYS_ATTRIBUTES 209
+#define KVM_CAP_PMU_CAPABILITY 210
#ifdef KVM_CAP_IRQ_ROUTING
@@ -1970,6 +1971,8 @@ struct kvm_dirty_gfn {
#define KVM_BUS_LOCK_DETECTION_OFF (1 << 0)
#define KVM_BUS_LOCK_DETECTION_EXIT (1 << 1)
+#define KVM_CAP_PMU_DISABLE (1 << 0)
+
/**
* struct kvm_stats_header - Header of per vm/vcpu binary statistics data.
* @flags: Some extra information for header, always 0 for now.
diff --git a/tools/include/uapi/linux/kvm.h b/tools/include/uapi/linux/kvm.h
index 5191b57e1562..3d6046295755 100644
--- a/tools/include/uapi/linux/kvm.h
+++ b/tools/include/uapi/linux/kvm.h
@@ -1134,6 +1134,7 @@ struct kvm_ppc_resize_hpt {
#define KVM_CAP_VM_GPA_BITS 207
#define KVM_CAP_XSAVE2 208
#define KVM_CAP_SYS_ATTRIBUTES 209
+#define KVM_CAP_PMU_CAPABILITY 210
#ifdef KVM_CAP_IRQ_ROUTING
@@ -1970,6 +1971,8 @@ struct kvm_dirty_gfn {
#define KVM_BUS_LOCK_DETECTION_OFF (1 << 0)
#define KVM_BUS_LOCK_DETECTION_EXIT (1 << 1)
+#define KVM_CAP_PMU_DISABLE (1 << 0)
+
/**
* struct kvm_stats_header - Header of per vm/vcpu binary statistics data.
* @flags: Some extra information for header, always 0 for now.
--
2.35.1.265.g69c8d7142f-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v7 2/3] KVM: selftests: Carve out helper to create "default" VM without vCPUs
2022-02-15 1:48 [PATCH v7 0/3] KVM: x86: Provide per VM capability for disabling PMU virtualization David Dunn
2022-02-15 1:48 ` [PATCH v7 1/3] " David Dunn
@ 2022-02-15 1:48 ` David Dunn
2022-02-17 15:46 ` Sean Christopherson
2022-02-15 1:48 ` [PATCH v7 3/3] KVM: selftests: Verify disabling PMU virtualization via KVM_CAP_CONFIG_PMU David Dunn
2 siblings, 1 reply; 6+ messages in thread
From: David Dunn @ 2022-02-15 1:48 UTC (permalink / raw)
To: pbonzini; +Cc: seanjc, jmattson, like.xu.linux, kvm, David Dunn
Carve out portion of vm_create_default so that selftests can modify
a "default" VM prior to creating vcpus.
Signed-off-by: David Dunn <daviddunn@google.com>
---
.../selftests/kvm/include/kvm_util_base.h | 3 ++
tools/testing/selftests/kvm/lib/kvm_util.c | 35 +++++++++++++++----
2 files changed, 32 insertions(+), 6 deletions(-)
diff --git a/tools/testing/selftests/kvm/include/kvm_util_base.h b/tools/testing/selftests/kvm/include/kvm_util_base.h
index 4ed6aa049a91..f987cf7c0d2e 100644
--- a/tools/testing/selftests/kvm/include/kvm_util_base.h
+++ b/tools/testing/selftests/kvm/include/kvm_util_base.h
@@ -336,6 +336,9 @@ struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
uint32_t num_percpu_pages, void *guest_code,
uint32_t vcpuids[]);
+/* Create a default VM without any vcpus. */
+struct kvm_vm *vm_create_without_vcpus(enum vm_guest_mode mode, uint64_t pages);
+
/*
* Adds a vCPU with reasonable defaults (e.g. a stack)
*
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index d8cf851ab119..5aea7734cfe3 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -362,6 +362,34 @@ struct kvm_vm *vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
return vm;
}
+/*
+ * Default VM creation without creating VCPUs
+ *
+ * Input Args:
+ * mode - VM Mode (e.g. VM_MODE_P52V48_4K)
+ * pages - pages of memory required for VM
+ *
+ * Output Args: None
+ *
+ * Return:
+ * Pointer to opaque structure that describes the created VM.
+ *
+ * Creates a VM with the mode specified by mode (e.g. VM_MODE_P52V48_4K).
+ */
+struct kvm_vm *vm_create_without_vcpus(enum vm_guest_mode mode, uint64_t pages)
+{
+ struct kvm_vm *vm;
+
+ vm = vm_create(mode, pages, O_RDWR);
+
+ kvm_vm_elf_load(vm, program_invocation_name);
+
+#ifdef __x86_64__
+ vm_create_irqchip(vm);
+#endif
+ return vm;
+}
+
/*
* VM Create with customized parameters
*
@@ -412,13 +440,8 @@ struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
nr_vcpus, kvm_check_cap(KVM_CAP_MAX_VCPUS));
pages = vm_adjust_num_guest_pages(mode, pages);
- vm = vm_create(mode, pages, O_RDWR);
-
- kvm_vm_elf_load(vm, program_invocation_name);
-#ifdef __x86_64__
- vm_create_irqchip(vm);
-#endif
+ vm = vm_create_without_vcpus(mode, pages);
for (i = 0; i < nr_vcpus; ++i) {
uint32_t vcpuid = vcpuids ? vcpuids[i] : i;
--
2.35.1.265.g69c8d7142f-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v7 3/3] KVM: selftests: Verify disabling PMU virtualization via KVM_CAP_CONFIG_PMU
2022-02-15 1:48 [PATCH v7 0/3] KVM: x86: Provide per VM capability for disabling PMU virtualization David Dunn
2022-02-15 1:48 ` [PATCH v7 1/3] " David Dunn
2022-02-15 1:48 ` [PATCH v7 2/3] KVM: selftests: Carve out helper to create "default" VM without vCPUs David Dunn
@ 2022-02-15 1:48 ` David Dunn
2 siblings, 0 replies; 6+ messages in thread
From: David Dunn @ 2022-02-15 1:48 UTC (permalink / raw)
To: pbonzini; +Cc: seanjc, jmattson, like.xu.linux, kvm, David Dunn
On a VM with PMU disabled via KVM_CAP_PMU_CONFIG, the PMU should not be
usable by the guest.
Signed-off-by: David Dunn <daviddunn@google.com>
---
.../kvm/x86_64/pmu_event_filter_test.c | 33 +++++++++++++++++++
1 file changed, 33 insertions(+)
diff --git a/tools/testing/selftests/kvm/x86_64/pmu_event_filter_test.c b/tools/testing/selftests/kvm/x86_64/pmu_event_filter_test.c
index c715adcbd487..dc04a266b3ed 100644
--- a/tools/testing/selftests/kvm/x86_64/pmu_event_filter_test.c
+++ b/tools/testing/selftests/kvm/x86_64/pmu_event_filter_test.c
@@ -325,6 +325,37 @@ static void test_not_member_allow_list(struct kvm_vm *vm)
TEST_ASSERT(!count, "Disallowed PMU Event is counting");
}
+/*
+ * Verify KVM_CAP_PMU_DISABLE prevents the use of the PMU.
+ *
+ * Note that KVM_CAP_PMU_CAPABILITY must be invoked prior to creating VCPUs.
+ */
+static void test_pmu_config_disable(void (*guest_code)(void))
+{
+ int r;
+ struct kvm_vm *vm;
+ struct kvm_enable_cap cap = { 0 };
+
+ r = kvm_check_cap(KVM_CAP_PMU_CAPABILITY);
+ if (!(r & KVM_CAP_PMU_DISABLE))
+ return;
+
+ vm = vm_create_without_vcpus(VM_MODE_DEFAULT, DEFAULT_GUEST_PHY_PAGES);
+
+ cap.cap = KVM_CAP_PMU_CAPABILITY;
+ cap.args[0] = KVM_CAP_PMU_DISABLE;
+ TEST_ASSERT(!vm_enable_cap(vm, &cap), "Failed KVM_CAP_PMU_DISABLE.");
+
+ vm_vcpu_add_default(vm, VCPU_ID, guest_code);
+ vm_init_descriptor_tables(vm);
+ vcpu_init_descriptor_tables(vm, VCPU_ID);
+
+ TEST_ASSERT(!sanity_check_pmu(vm),
+ "Guest should not be able to use disabled PMU.");
+
+ kvm_vm_free(vm);
+}
+
/*
* Check for a non-zero PMU version, at least one general-purpose
* counter per logical processor, an EBX bit vector of length greater
@@ -430,5 +461,7 @@ int main(int argc, char *argv[])
kvm_vm_free(vm);
+ test_pmu_config_disable(guest_code);
+
return 0;
}
--
2.35.1.265.g69c8d7142f-goog
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v7 2/3] KVM: selftests: Carve out helper to create "default" VM without vCPUs
2022-02-15 1:48 ` [PATCH v7 2/3] KVM: selftests: Carve out helper to create "default" VM without vCPUs David Dunn
@ 2022-02-17 15:46 ` Sean Christopherson
0 siblings, 0 replies; 6+ messages in thread
From: Sean Christopherson @ 2022-02-17 15:46 UTC (permalink / raw)
To: David Dunn; +Cc: pbonzini, jmattson, like.xu.linux, kvm
On Tue, Feb 15, 2022, David Dunn wrote:
> Carve out portion of vm_create_default so that selftests can modify
> a "default" VM prior to creating vcpus.
>
> Signed-off-by: David Dunn <daviddunn@google.com>
> ---
> .../selftests/kvm/include/kvm_util_base.h | 3 ++
> tools/testing/selftests/kvm/lib/kvm_util.c | 35 +++++++++++++++----
> 2 files changed, 32 insertions(+), 6 deletions(-)
>
> diff --git a/tools/testing/selftests/kvm/include/kvm_util_base.h b/tools/testing/selftests/kvm/include/kvm_util_base.h
> index 4ed6aa049a91..f987cf7c0d2e 100644
> --- a/tools/testing/selftests/kvm/include/kvm_util_base.h
> +++ b/tools/testing/selftests/kvm/include/kvm_util_base.h
> @@ -336,6 +336,9 @@ struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
> uint32_t num_percpu_pages, void *guest_code,
> uint32_t vcpuids[]);
>
> +/* Create a default VM without any vcpus. */
> +struct kvm_vm *vm_create_without_vcpus(enum vm_guest_mode mode, uint64_t pages);
> +
> /*
> * Adds a vCPU with reasonable defaults (e.g. a stack)
> *
> diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
> index d8cf851ab119..5aea7734cfe3 100644
> --- a/tools/testing/selftests/kvm/lib/kvm_util.c
> +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
> @@ -362,6 +362,34 @@ struct kvm_vm *vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
> return vm;
> }
>
> +/*
> + * Default VM creation without creating VCPUs
> + *
> + * Input Args:
> + * mode - VM Mode (e.g. VM_MODE_P52V48_4K)
> + * pages - pages of memory required for VM
> + *
> + * Output Args: None
> + *
> + * Return:
> + * Pointer to opaque structure that describes the created VM.
> + *
> + * Creates a VM with the mode specified by mode (e.g. VM_MODE_P52V48_4K).
> + */
I vote to omit this function comment. Most of the existing comments in kvm_util.c
are a waste of space, and (no offence) this is no different. And I'm planning on
deleting most of the existing boilerplate comments that don't provide any insight :-)
> +struct kvm_vm *vm_create_without_vcpus(enum vm_guest_mode mode, uint64_t pages)
> +{
> + struct kvm_vm *vm;
> +
> + vm = vm_create(mode, pages, O_RDWR);
> +
> + kvm_vm_elf_load(vm, program_invocation_name);
> +
> +#ifdef __x86_64__
> + vm_create_irqchip(vm);
> +#endif
> + return vm;
> +}
> +
> /*
> * VM Create with customized parameters
> *
> @@ -412,13 +440,8 @@ struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
> nr_vcpus, kvm_check_cap(KVM_CAP_MAX_VCPUS));
>
> pages = vm_adjust_num_guest_pages(mode, pages);
> - vm = vm_create(mode, pages, O_RDWR);
> -
> - kvm_vm_elf_load(vm, program_invocation_name);
>
> -#ifdef __x86_64__
> - vm_create_irqchip(vm);
> -#endif
> + vm = vm_create_without_vcpus(mode, pages);
>
> for (i = 0; i < nr_vcpus; ++i) {
> uint32_t vcpuid = vcpuids ? vcpuids[i] : i;
> --
> 2.35.1.265.g69c8d7142f-goog
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v7 1/3] KVM: x86: Provide per VM capability for disabling PMU virtualization
2022-02-15 1:48 ` [PATCH v7 1/3] " David Dunn
@ 2022-02-17 16:22 ` Sean Christopherson
0 siblings, 0 replies; 6+ messages in thread
From: Sean Christopherson @ 2022-02-17 16:22 UTC (permalink / raw)
To: David Dunn; +Cc: pbonzini, jmattson, like.xu.linux, kvm
On Tue, Feb 15, 2022, David Dunn wrote:
> Add a new capability, KVM_CAP_PMU_CAPABILITY, that takes a bitmask of
> settings/features to allow userspace to configure PMU virtualization on
> a per-VM basis. For now, support a single flag, KVM_CAP_PMU_DISABLE,
> to allow disabling PMU virtualization for a VM even when KVM is configured
> with enable_pmu=true a module level.
>
> To keep KVM simple, disallow changing VM's PMU configuration after vCPUs
> have been created.
>
> Signed-off-by: David Dunn <daviddunn@google.com>
> ---
A few nits, otherwise
Reviewed-by: Sean Christopherson <seanjc@google.com>
Somewhat off-topic, looking at this got me looking at vmx_get_perf_capabilities().
We really should cache the result of its RDMSR.
And I believe emulation of WRMSR from the guest to MSR_IA32_PERF_CAPABILITIES is
missing a guest_cpuid_has() check.
> Documentation/virt/kvm/api.rst | 22 ++++++++++++++++++++++
> arch/x86/include/asm/kvm_host.h | 1 +
> arch/x86/kvm/svm/pmu.c | 2 +-
> arch/x86/kvm/vmx/pmu_intel.c | 2 +-
> arch/x86/kvm/x86.c | 17 +++++++++++++++++
> include/uapi/linux/kvm.h | 3 +++
> tools/include/uapi/linux/kvm.h | 3 +++
> 7 files changed, 48 insertions(+), 2 deletions(-)
>
> diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
> index a4267104db50..df836965b347 100644
> --- a/Documentation/virt/kvm/api.rst
> +++ b/Documentation/virt/kvm/api.rst
> @@ -7561,3 +7561,25 @@ The argument to KVM_ENABLE_CAP is also a bitmask, and must be a subset
> of the result of KVM_CHECK_EXTENSION. KVM will forward to userspace
> the hypercalls whose corresponding bit is in the argument, and return
> ENOSYS for the others.
> +
> +8.35 KVM_CAP_PMU_CAPABILITY
> +---------------------------
> +
> +:Capability KVM_CAP_PMU_CAPABILITY
> +:Architectures: x86
> +:Type: vm
> +:Parameters: arg[0] is bitmask of PMU virtualization capabilities.
Referring to "DISABLE" as a capabilitiy is rather odd, but I don't have a better
suggestion.
> +:Returns 0 on success, -EINVAL when arg[0] contains invalid bits
> +
> +This capability alters PMU virtualization in KVM.
> +
> +Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of
> +PMU virtualization capabilities that can be adjusted on a VM.
> +
> +The argument to KVM_ENABLE_CAP is also a bitmask and selects specific
> +PMU virtualization capabilities to be applied to the VM. This can
> +only be invoked on a VM prior to the creation of VCPUs.
> +
> +At this time, KVM_CAP_PMU_DISABLE is the only capability. Setting
> +this capability will disable PMU virtualization for that VM. Usermode
> +should adjust CPUID leaf 0xA to reflect that the PMU is disabled.
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 10815b672a26..61e9050a3488 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -1233,6 +1233,7 @@ struct kvm_arch {
> hpa_t hv_root_tdp;
> spinlock_t hv_root_tdp_lock;
> #endif
> + bool enable_pmu;
Doesn''t really matter, but I'd prefer we put this up among the many other bools, e.g.
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 4bb0d3be2055..b57d40419ddd 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1150,6 +1150,7 @@ struct kvm_arch {
bool guest_can_read_msr_platform_info;
bool exception_payload_enabled;
+ bool enable_pmu;
bool bus_lock_detection_enabled;
/*
* If exit_on_emulation_error is set, and the in-kernel instruction
@@ -1233,7 +1234,6 @@ struct kvm_arch {
hpa_t hv_root_tdp;
spinlock_t hv_root_tdp_lock;
#endif
- bool enable_pmu;
};
struct kvm_vm_stat {
> };
>
> struct kvm_vm_stat {
> diff --git a/arch/x86/kvm/svm/pmu.c b/arch/x86/kvm/svm/pmu.c
> index 5aa45f13b16d..d4de52409335 100644
> --- a/arch/x86/kvm/svm/pmu.c
> +++ b/arch/x86/kvm/svm/pmu.c
> @@ -101,7 +101,7 @@ static inline struct kvm_pmc *get_gp_pmc_amd(struct kvm_pmu *pmu, u32 msr,
> {
> struct kvm_vcpu *vcpu = pmu_to_vcpu(pmu);
>
> - if (!enable_pmu)
> + if (!vcpu->kvm->arch.enable_pmu)
> return NULL;
>
> switch (msr) {
> diff --git a/arch/x86/kvm/vmx/pmu_intel.c b/arch/x86/kvm/vmx/pmu_intel.c
> index 03fab48b149c..4e5b1eeeb77c 100644
> --- a/arch/x86/kvm/vmx/pmu_intel.c
> +++ b/arch/x86/kvm/vmx/pmu_intel.c
> @@ -487,7 +487,7 @@ static void intel_pmu_refresh(struct kvm_vcpu *vcpu)
> pmu->reserved_bits = 0xffffffff00200000ull;
>
> entry = kvm_find_cpuid_entry(vcpu, 0xa, 0);
> - if (!entry || !enable_pmu)
> + if (!entry || !vcpu->kvm->arch.enable_pmu)
> return;
> eax.full = entry->eax;
> edx.full = entry->edx;
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index eaa3b5b89c5e..0803a1388bbf 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -110,6 +110,8 @@ static u64 __read_mostly cr4_reserved_bits = CR4_RESERVED_BITS;
>
> #define KVM_EXIT_HYPERCALL_VALID_MASK (1 << KVM_HC_MAP_GPA_RANGE)
>
> +#define KVM_CAP_PMU_VALID_MASK KVM_CAP_PMU_DISABLE
> +
> #define KVM_X2APIC_API_VALID_FLAGS (KVM_X2APIC_API_USE_32BIT_IDS | \
> KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK)
>
> @@ -4331,6 +4333,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
> if (r < sizeof(struct kvm_xsave))
> r = sizeof(struct kvm_xsave);
> break;
> + case KVM_CAP_PMU_CAPABILITY:
> + r = enable_pmu ? KVM_CAP_PMU_VALID_MASK : 0;
> + break;
> }
> default:
> break;
> @@ -6005,6 +6010,17 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
> kvm->arch.exit_on_emulation_error = cap->args[0];
> r = 0;
> break;
> + case KVM_CAP_PMU_CAPABILITY:
> + r = -EINVAL;
> + if (!enable_pmu || (cap->args[0] & ~KVM_CAP_PMU_VALID_MASK))
> + break;
I'd prefer another newline here, the other case statements are just bad influences :-)
> + mutex_lock(&kvm->lock);
> + if (!kvm->created_vcpus) {
> + kvm->arch.enable_pmu = !(cap->args[0] & KVM_CAP_PMU_DISABLE);
> + r = 0;
> + }
> + mutex_unlock(&kvm->lock);
> + break;
> default:
> r = -EINVAL;
> break;
^ permalink raw reply related [flat|nested] 6+ messages in thread
end of thread, other threads:[~2022-02-17 16:22 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-02-15 1:48 [PATCH v7 0/3] KVM: x86: Provide per VM capability for disabling PMU virtualization David Dunn
2022-02-15 1:48 ` [PATCH v7 1/3] " David Dunn
2022-02-17 16:22 ` Sean Christopherson
2022-02-15 1:48 ` [PATCH v7 2/3] KVM: selftests: Carve out helper to create "default" VM without vCPUs David Dunn
2022-02-17 15:46 ` Sean Christopherson
2022-02-15 1:48 ` [PATCH v7 3/3] KVM: selftests: Verify disabling PMU virtualization via KVM_CAP_CONFIG_PMU David Dunn
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.