Skip to contentSkip to conversation

arm64: Support perf event modifiers :G and :H

PATCH Superseded Andrew Murray 3 participants 22 messages Mar 28, 2019 – Apr 9, 2019

Subsystems (6)

v12 of 13 Superseded Latest is v13 →

lore
Share & tools
lore
#0 lore raw

v12 cover arm64: Support perf event modifiers :G and :H

This patchset provides support for perf event modifiers :G and :H which allows for filtering of PMU events between host and guests when used with KVM.

As the underlying hardware cannot distinguish between guest and host context we must switch the performance counters upon entry/exit to the guest.

For non-VHE the counters must be stopped and started upon entry/exit to the guest in __kvm_vcpu_run_nvhe.

For VHE we keep the counters enabled but instead change the event type to include/exclude EL0 as appropriate. This allows us to perform this switch outside the critical section of world-switch code in vcpu_load.

This has been tested with VHE and non-VHE kernels with a KVM guest.

Changes from v11:

- Rebased to v5.1-rc2

- Minor changes to commit messages, fixing typos, removing superfluous comments, code.

- Change ^ to != in kvm_pmu_switch_needed

Changes from v10:

- Remove counter switch code from kvm_vcpu_run_vhe and replace with enable/disable EL0 event type in kvm_arch_vcpu_load instead

- Reduce counter switch code in __kvm_vcpu_run_nvhe by moving some logic to kvm_set_pmu_events (kvm_pmu_switch_needed)

- Simplify code by removing need for KVM_PMU_EVENTS_{HOST,GUEST}

- Add kvm_host helper function to let PMU determine if event should start counting when it is enabled

- Exclude EL2 on !VHE when exclude_host (to avoid counting host events as guest during entry/exit)

- Moved PMU switching code to its own file

- Rebased to v5.0

- Added documentation

- Removed Reviewed-By's for changed patches and updated commit messages

Changes from v9:

- Rebased to v5.0-rc2
- Ensure get_host_ctxt considers host_ctxt offset

Changes from v8:

- Added additional comments
- Fixed bisect build failure
- Renamed cpu_ctxt variable to cpu_data

Changes from v7:

- Added additional patch to encapsulate kvm_cpu_context in kvm_host_data

Changes from v6:

- Move events_host/events_guest out of kvm_cpu_context

Changes from v5:

- Tweak logic in use of kvm_set_pmu_events

Changes from v4:

- Prevent unnecessary write_sysreg calls by improving __pmu_switch_to_xxx logic.

Changes from v3:

- Remove confusing _only suffix from bitfields in kvm_cpu_context
- Remove unnecessary condition when clearing event bits in disable
- Simplify API of KVM accessors
- Prevent unnecessary setting of pmcnten when guest/host events are the same.

Changes from v2:

- Ensured that exclude_kernel works for guest
- Removed unnecessary exclusion of EL2 with exclude_host on !VHE
- Renamed kvm_clr_set_host_pmu_events to reflect args order
- Added additional information to isb patch

Changes from v1:

- Removed unnecessary exclusion of EL1 with exclude_guest on VHE
- Removed unnecessary isb from existing perf_event.c driver
- Folded perf_event.c patches together
- Added additional information to last patch commit message

Andrew Murray (8):
arm64: arm_pmu: remove unnecessary isb instruction
arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data
arm64: KVM: add accessors to track guest/host only counters
arm64: arm_pmu: Add !VHE support for exclude_host/exclude_guest
attributes
arm64: KVM: Enable !VHE support for :G/:H perf event modifiers
arm64: KVM: Enable VHE support for :G/:H perf event modifiers
arm64: KVM: avoid isb's by using direct pmxevtyper sysreg
arm64: docs: document perf event attributes

 Documentation/arm64/perf.txt      |  74 ++++++++++
 arch/arm/include/asm/kvm_host.h   |  11 +-
 arch/arm64/include/asm/kvm_asm.h  |   3 +-
 arch/arm64/include/asm/kvm_host.h |  39 ++++-
 arch/arm64/kernel/asm-offsets.c   |   1 +
 arch/arm64/kernel/perf_event.c    |  50 +++++--
 arch/arm64/kvm/Makefile           |   2 +-
 arch/arm64/kvm/hyp/switch.c       |   6 +
 arch/arm64/kvm/pmu.c              | 232 ++++++++++++++++++++++++++++++
 arch/arm64/kvm/sys_regs.c         |   3 +
 virt/kvm/arm/arm.c                |  16 ++-
 11 files changed, 412 insertions(+), 25 deletions(-)
 create mode 100644 Documentation/arm64/perf.txt
 create mode 100644 arch/arm64/kvm/pmu.c
Signature · 1 line
2.21.0
#1 lore raw

PATCH v12 1/8 arm64: arm_pmu: remove unnecessary isb instruction

The armv8pmu_enable_event_counter function issues an isb instruction after enabling a pair of counters - this doesn't provide any value and is inconsistent with the armv8pmu_disable_event_counter.

In any case armv8pmu_enable_event_counter is always called with the PMU stopped. Starting the PMU with armv8pmu_start results in an isb instruction being issued prior to writing to PMCR_EL0.

Let's remove the unnecessary isb instruction.

---
 arch/arm64/kernel/perf_event.c | 1 -
 1 file changed, 1 deletion(-)
arch/arm64/kernel/perf_event.c +0 -1
index 4addb38bc250..cccf4fc86d67 100644
@@ -533,7 +533,6 @@ static inline void armv8pmu_enable_event_counter(struct perf_event *event)
 	armv8pmu_enable_counter(idx);
 	if (armv8pmu_event_is_chained(event))
 		armv8pmu_enable_counter(idx - 1);
-	isb();
 }
 
 static inline int armv8pmu_disable_counter(int idx)
Signature · 1 line
2.21.0
#2 lore raw

PATCH v12 2/8 arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data

In reply to Andrew Murray

The virt/arm core allocates a kvm_cpu_context_t percpu, at present this is a typedef to kvm_cpu_context and is used to store host cpu context. The kvm_cpu_context structure is also used elsewhere to hold vcpu context. In order to use the percpu to hold additional future host information we encapsulate kvm_cpu_context in a new structure and rename the typedef and percpu to match.

---
 arch/arm/include/asm/kvm_host.h   |  8 ++++++--
 arch/arm64/include/asm/kvm_asm.h  |  3 ++-
 arch/arm64/include/asm/kvm_host.h | 16 ++++++++++------
 arch/arm64/kernel/asm-offsets.c   |  1 +
 virt/kvm/arm/arm.c                | 14 ++++++++------
 5 files changed, 27 insertions(+), 15 deletions(-)
arch/arm/include/asm/kvm_host.h +6 -2
index 770d73257ad9..427c28be6452 100644
@@ -150,7 +150,11 @@ struct kvm_cpu_context {
 	u32 cp15[NR_CP15_REGS];
 };
 
-typedef struct kvm_cpu_context kvm_cpu_context_t;
+struct kvm_host_data {
+	struct kvm_cpu_context host_ctxt;
+};
+
+typedef struct kvm_host_data kvm_host_data_t;
 
 static inline void kvm_init_host_cpu_context(kvm_cpu_context_t *cpu_ctxt,
 					     int cpu)
@@ -182,7 +186,7 @@ struct kvm_vcpu_arch {
 	struct kvm_vcpu_fault_info fault;
 
 	/* Host FP context */
-	kvm_cpu_context_t *host_cpu_context;
+	struct kvm_cpu_context *host_cpu_context;
 
 	/* VGIC state */
 	struct vgic_cpu vgic_cpu;
arch/arm64/include/asm/kvm_asm.h +2 -1
index f5b79e995f40..ff73f5462aca 100644
@@ -108,7 +108,8 @@ extern u32 __kvm_get_mdcr_el2(void);
 .endm
 
 .macro get_host_ctxt reg, tmp
-	hyp_adr_this_cpu \reg, kvm_host_cpu_state, \tmp
+	hyp_adr_this_cpu \reg, kvm_host_data, \tmp
+	add	\reg, \reg, #HOST_DATA_CONTEXT
 .endm
 
 .macro get_vcpu_ptr vcpu, ctxt
arch/arm64/include/asm/kvm_host.h +10 -6
index a01fe087e022..6530ad522a16 100644
@@ -212,7 +212,11 @@ struct kvm_cpu_context {
 	struct kvm_vcpu *__hyp_running_vcpu;
 };
 
-typedef struct kvm_cpu_context kvm_cpu_context_t;
+struct kvm_host_data {
+	struct kvm_cpu_context host_ctxt;
+};
+
+typedef struct kvm_host_data kvm_host_data_t;
 
 struct vcpu_reset_state {
 	unsigned long	pc;
@@ -255,7 +259,7 @@ struct kvm_vcpu_arch {
 	struct kvm_guest_debug_arch external_debug_state;
 
 	/* Pointer to host CPU context */
-	kvm_cpu_context_t *host_cpu_context;
+	struct kvm_cpu_context *host_cpu_context;
 
 	struct thread_info *host_thread_info;	/* hyp VA */
 	struct user_fpsimd_state *host_fpsimd_state;	/* hyp VA */
@@ -432,9 +436,9 @@ void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
 
 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
 
-DECLARE_PER_CPU(kvm_cpu_context_t, kvm_host_cpu_state);
+DECLARE_PER_CPU(kvm_host_data_t, kvm_host_data);
 
-static inline void kvm_init_host_cpu_context(kvm_cpu_context_t *cpu_ctxt,
+static inline void kvm_init_host_cpu_context(struct kvm_cpu_context *cpu_ctxt,
 					     int cpu)
 {
 	/* The host's MPIDR is immutable, so let's set it up at boot time */
@@ -452,8 +456,8 @@ static inline void __cpu_init_hyp_mode(phys_addr_t pgd_ptr,
 	 * kernel's mapping to the linear mapping, and store it in tpidr_el2
 	 * so that we can use adr_l to access per-cpu variables in EL2.
 	 */
-	u64 tpidr_el2 = ((u64)this_cpu_ptr(&kvm_host_cpu_state) -
-			 (u64)kvm_ksym_ref(kvm_host_cpu_state));
+	u64 tpidr_el2 = ((u64)this_cpu_ptr(&kvm_host_data) -
+			 (u64)kvm_ksym_ref(kvm_host_data));
 
 	/*
 	 * Call initialization code, and switch to the full blown HYP code.
arch/arm64/kernel/asm-offsets.c +1 -0
index 7f40dcbdd51d..fbf0409cc2be 100644
@@ -128,6 +128,7 @@ int main(void)
   DEFINE(CPU_GP_REGS,		offsetof(struct kvm_cpu_context, gp_regs));
   DEFINE(CPU_USER_PT_REGS,	offsetof(struct kvm_regs, regs));
   DEFINE(HOST_CONTEXT_VCPU,	offsetof(struct kvm_cpu_context, __hyp_running_vcpu));
+  DEFINE(HOST_DATA_CONTEXT,	offsetof(struct kvm_host_data, host_ctxt));
 #endif
 #ifdef CONFIG_CPU_PM
   DEFINE(CPU_CTX_SP,		offsetof(struct cpu_suspend_ctx, sp));
virt/kvm/arm/arm.c +8 -6
index 99c37384ba7b..6bcda52c0ea6 100644
@@ -56,7 +56,7 @@
 __asm__(".arch_extension	virt");
 #endif
 
-DEFINE_PER_CPU(kvm_cpu_context_t, kvm_host_cpu_state);
+DEFINE_PER_CPU(kvm_host_data_t, kvm_host_data);
 static DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);
 
 /* Per-CPU variable containing the currently running vcpu. */
@@ -360,8 +360,10 @@ int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 {
 	int *last_ran;
+	kvm_host_data_t *cpu_data;
 
 	last_ran = this_cpu_ptr(vcpu->kvm->arch.last_vcpu_ran);
+	cpu_data = this_cpu_ptr(&kvm_host_data);
 
 	/*
 	 * We might get preempted before the vCPU actually runs, but
@@ -373,7 +375,7 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 	}
 
 	vcpu->cpu = cpu;
-	vcpu->arch.host_cpu_context = this_cpu_ptr(&kvm_host_cpu_state);
+	vcpu->arch.host_cpu_context = &cpu_data->host_ctxt;
 
 	kvm_arm_set_running_vcpu(vcpu);
 	kvm_vgic_load(vcpu);
@@ -1549,11 +1551,11 @@ static int init_hyp_mode(void)
 	}
 
 	for_each_possible_cpu(cpu) {
-		kvm_cpu_context_t *cpu_ctxt;
+		kvm_host_data_t *cpu_data;
 
-		cpu_ctxt = per_cpu_ptr(&kvm_host_cpu_state, cpu);
-		kvm_init_host_cpu_context(cpu_ctxt, cpu);
-		err = create_hyp_mappings(cpu_ctxt, cpu_ctxt + 1, PAGE_HYP);
+		cpu_data = per_cpu_ptr(&kvm_host_data, cpu);
+		kvm_init_host_cpu_context(&cpu_data->host_ctxt, cpu);
+		err = create_hyp_mappings(cpu_data, cpu_data + 1, PAGE_HYP);
 
 		if (err) {
 			kvm_err("Cannot map host CPU state: %d\n", err);
Signature · 1 line
2.21.0
reviewer Reviewed-by on v12 2/8
#3 lore raw

Re: [PATCH v12 2/8] arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data

Quoted 23 lines · Andrew Murray

The virt/arm core allocates a kvm_cpu_context_t percpu, at present this is a typedef to kvm_cpu_context and is used to store host cpu context. The kvm_cpu_context structure is also used elsewhere to hold vcpu context. In order to use the percpu to hold additional future host information we encapsulate kvm_cpu_context in a new structure and rename the typedef and percpu to match.

---
  arch/arm/include/asm/kvm_host.h   |  8 ++++++--
  arch/arm64/include/asm/kvm_asm.h  |  3 ++-
  arch/arm64/include/asm/kvm_host.h | 16 ++++++++++------
  arch/arm64/kernel/asm-offsets.c   |  1 +
  virt/kvm/arm/arm.c                | 14 ++++++++------
  5 files changed, 27 insertions(+), 15 deletions(-)
arch/arm/include/asm/kvm_host.h +0 -0
index 770d73257ad9..427c28be6452 100644
@@ -150,7 +150,11 @@ struct kvm_cpu_context {
  	u32 cp15[NR_CP15_REGS];
  };

arch/arm/include/asm/kvm_host.h struct kvm_cpu_context { · line 160

Quoted 10 lines
arch/arm/include/asm/kvm_host.h +5 -1
 
-typedef struct kvm_cpu_context kvm_cpu_context_t;
+struct kvm_host_data {
+	struct kvm_cpu_context host_ctxt;
+};
+
+typedef struct kvm_host_data kvm_host_data_t;
 
  static inline void kvm_init_host_cpu_context(kvm_cpu_context_t *cpu_ctxt,
  					     int cpu)

We need to fix this function prototype to accept struct kvm_cpu_context, instead of the now removed kvm_cpu_context_t, to prevent a build break on arm32 ?

With that :

#4 lore raw

Re: [PATCH v12 2/8] arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data

Quoted 28 lines · Suzuki K Poulose
Quoted 27 lines · Andrew Murray

The virt/arm core allocates a kvm_cpu_context_t percpu, at present this is a typedef to kvm_cpu_context and is used to store host cpu context. The kvm_cpu_context structure is also used elsewhere to hold vcpu context. In order to use the percpu to hold additional future host information we encapsulate kvm_cpu_context in a new structure and rename the typedef and percpu to match.

---
  arch/arm/include/asm/kvm_host.h   |  8 ++++++--
  arch/arm64/include/asm/kvm_asm.h  |  3 ++-
  arch/arm64/include/asm/kvm_host.h | 16 ++++++++++------
  arch/arm64/kernel/asm-offsets.c   |  1 +
  virt/kvm/arm/arm.c                | 14 ++++++++------
  5 files changed, 27 insertions(+), 15 deletions(-)
arch/arm/include/asm/kvm_host.h +3 -1
index 770d73257ad9..427c28be6452 100644
@@ -150,7 +150,11 @@ struct kvm_cpu_context {
  	u32 cp15[NR_CP15_REGS];
  };
-typedef struct kvm_cpu_context kvm_cpu_context_t;
+struct kvm_host_data {
+	struct kvm_cpu_context host_ctxt;
+};
Quoted 8 lines

arch/arm/include/asm/kvm_host.h struct kvm_cpu_context { · line 160

Quoted 4 lines
arch/arm/include/asm/kvm_host.h +2 -0
+
+typedef struct kvm_host_data kvm_host_data_t;
  static inline void kvm_init_host_cpu_context(kvm_cpu_context_t *cpu_ctxt,
  					     int cpu)

We need to fix this function prototype to accept struct kvm_cpu_context, instead of the now removed kvm_cpu_context_t, to prevent a build break on arm32 ?

Yes this was a breakage, thanks for pointing this out.

Thanks,

Andrew Murray

Quoted 4 lines

With that :

#5 lore raw

Re: [PATCH v12 2/8] arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data

In reply to Andrew Murray
Quoted 7 lines · Andrew Murray

The virt/arm core allocates a kvm_cpu_context_t percpu, at present this is a typedef to kvm_cpu_context and is used to store host cpu context. The kvm_cpu_context structure is also used elsewhere to hold vcpu context. In order to use the percpu to hold additional future host information we encapsulate kvm_cpu_context in a new structure and rename the typedef and percpu to match.

Quoted 8 lines
---
 arch/arm/include/asm/kvm_host.h   |  8 ++++++--
 arch/arm64/include/asm/kvm_asm.h  |  3 ++-
 arch/arm64/include/asm/kvm_host.h | 16 ++++++++++------
 arch/arm64/kernel/asm-offsets.c   |  1 +
 virt/kvm/arm/arm.c                | 14 ++++++++------
 5 files changed, 27 insertions(+), 15 deletions(-)

As per usual, I'll need an Ack from Christoffer and/or Marc on this one :)

Will

#6 lore raw

Re: [PATCH v12 2/8] arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data

Quoted 17 lines · Will Deacon
Quoted 7 lines · Andrew Murray

The virt/arm core allocates a kvm_cpu_context_t percpu, at present this is a typedef to kvm_cpu_context and is used to store host cpu context. The kvm_cpu_context structure is also used elsewhere to hold vcpu context. In order to use the percpu to hold additional future host information we encapsulate kvm_cpu_context in a new structure and rename the typedef and percpu to match.

Quoted 8 lines
---
 arch/arm/include/asm/kvm_host.h   |  8 ++++++--
 arch/arm64/include/asm/kvm_asm.h  |  3 ++-
 arch/arm64/include/asm/kvm_host.h | 16 ++++++++++------
 arch/arm64/kernel/asm-offsets.c   |  1 +
 virt/kvm/arm/arm.c                | 14 ++++++++------
 5 files changed, 27 insertions(+), 15 deletions(-)
Quoted 1 line

As per usual, I'll need an Ack from Christoffer and/or Marc on this one :)

Actually... given that most of this series is now firmly in KVM territory, it probably makes more sense for the kvm/arm tree to take this on a separate branch, which I can also pull into the arm64 perf tree if I run into conflicts on the non-kvm parts.

Will

#7 lore raw

Re: [PATCH v12 2/8] arm64: KVM: encapsulate kvm_cpu_context in kvm_host_data

Quoted 18 lines · Will Deacon
Quoted 17 lines · Will Deacon
Quoted 15 lines · Andrew Murray

The virt/arm core allocates a kvm_cpu_context_t percpu, at present this is a typedef to kvm_cpu_context and is used to store host cpu context. The kvm_cpu_context structure is also used elsewhere to hold vcpu context. In order to use the percpu to hold additional future host information we encapsulate kvm_cpu_context in a new structure and rename the typedef and percpu to match.

---
 arch/arm/include/asm/kvm_host.h   |  8 ++++++--
 arch/arm64/include/asm/kvm_asm.h  |  3 ++-
 arch/arm64/include/asm/kvm_host.h | 16 ++++++++++------
 arch/arm64/kernel/asm-offsets.c   |  1 +
 virt/kvm/arm/arm.c                | 14 ++++++++------
 5 files changed, 27 insertions(+), 15 deletions(-)
Quoted 6 lines
Quoted 1 line

As per usual, I'll need an Ack from Christoffer and/or Marc on this one :)

Actually... given that most of this series is now firmly in KVM territory, it probably makes more sense for the kvm/arm tree to take this on a separate branch, which I can also pull into the arm64 perf tree if I run into conflicts on the non-kvm parts.

No problem.

Thanks,

Andrew Murray

Quoted 2 lines

Will

#8 lore raw

PATCH v12 3/8 arm64: KVM: add accessors to track guest/host only counters

In reply to Andrew Murray

In order to effeciently switch events_{guest,host} perf counters at guest entry/exit we add bitfields to kvm_cpu_context for guest and host events as well as accessors for updating them.

A function is also provided which allows the PMU driver to determine if a counter should start counting when it is enabled. With exclude_host, we may only start counting when entering the guest.

---
 arch/arm64/include/asm/kvm_host.h | 17 ++++++++++++
 arch/arm64/kvm/Makefile           |  2 +-
 arch/arm64/kvm/pmu.c              | 45 +++++++++++++++++++++++++++++++
 3 files changed, 63 insertions(+), 1 deletion(-)
 create mode 100644 arch/arm64/kvm/pmu.c
arch/arm64/include/asm/kvm_host.h +17 -0
index 6530ad522a16..213465196e6b 100644
@@ -212,8 +212,14 @@ struct kvm_cpu_context {
 	struct kvm_vcpu *__hyp_running_vcpu;
 };
 
+struct kvm_pmu_events {
+	u32 events_host;
+	u32 events_guest;
+};
+
 struct kvm_host_data {
 	struct kvm_cpu_context host_ctxt;
+	struct kvm_pmu_events pmu_events;
 };
 
 typedef struct kvm_host_data kvm_host_data_t;
@@ -520,11 +526,22 @@ void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu);
 void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu);
 void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
 
+static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
+{
+	return attr->exclude_host;
+}
+
 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
 {
 	return kvm_arch_vcpu_run_map_fp(vcpu);
 }
+
+void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
+void kvm_clr_pmu_events(u32 clr);
+#else
+static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
+static inline void kvm_clr_pmu_events(u32 clr) {}
 #endif
 
 static inline void kvm_arm_vhe_guest_enter(void)
arch/arm64/kvm/Makefile +1 -1
index 690e033a91c0..3ac1a64d2fb9 100644
@@ -17,7 +17,7 @@ kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/arm/psci.o $(KVM)/arm/perf.o
 kvm-$(CONFIG_KVM_ARM_HOST) += inject_fault.o regmap.o va_layout.o
 kvm-$(CONFIG_KVM_ARM_HOST) += hyp.o hyp-init.o handle_exit.o
 kvm-$(CONFIG_KVM_ARM_HOST) += guest.o debug.o reset.o sys_regs.o sys_regs_generic_v8.o
-kvm-$(CONFIG_KVM_ARM_HOST) += vgic-sys-reg-v3.o fpsimd.o
+kvm-$(CONFIG_KVM_ARM_HOST) += vgic-sys-reg-v3.o fpsimd.o pmu.o
 kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/arm/aarch32.o
 
 kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/arm/vgic/vgic.o
arch/arm64/kvm/pmu.c new file +45 -0

45 lines not on this page · raw message

Signature · 1 line
2.21.0
#9 lore raw

PATCH v12 4/8 arm64: arm_pmu: Add !VHE support for exclude_host/exclude_guest attributes

In reply to Andrew Murray

Add support for the :G and :H attributes in perf by handling the exclude_host/exclude_guest event attributes.

We notify KVM of counters that we wish to be enabled or disabled on guest entry/exit and thus defer from starting or stopping events based on their event attributes.

With !VHE we switch the counters between host/guest at EL2. We are able to eliminate counters counting host events on the boundaries of guest entry/exit when using :G by filtering out EL2 for exclude_host. When using !exclude_hv there is a small blackout window at the guest entry/exit where host events are not captured.

---
 arch/arm64/kernel/perf_event.c | 43 ++++++++++++++++++++++++++++------
 1 file changed, 36 insertions(+), 7 deletions(-)
arch/arm64/kernel/perf_event.c +36 -7
index cccf4fc86d67..6bb28aaf5aea 100644
@@ -26,6 +26,7 @@
 
 #include <linux/acpi.h>
 #include <linux/clocksource.h>
+#include <linux/kvm_host.h>
 #include <linux/of.h>
 #include <linux/perf/arm_pmu.h>
 #include <linux/platform_device.h>
@@ -528,11 +529,21 @@ static inline int armv8pmu_enable_counter(int idx)
 
 static inline void armv8pmu_enable_event_counter(struct perf_event *event)
 {
+	struct perf_event_attr *attr = &event->attr;
 	int idx = event->hw.idx;
+	u32 counter_bits = BIT(ARMV8_IDX_TO_COUNTER(idx));
 
-	armv8pmu_enable_counter(idx);
 	if (armv8pmu_event_is_chained(event))
-		armv8pmu_enable_counter(idx - 1);
+		counter_bits |= BIT(ARMV8_IDX_TO_COUNTER(idx - 1));
+
+	kvm_set_pmu_events(counter_bits, attr);
+
+	/* We rely on the hypervisor switch code to enable guest counters */
+	if (!kvm_pmu_counter_deferred(attr)) {
+		armv8pmu_enable_counter(idx);
+		if (armv8pmu_event_is_chained(event))
+			armv8pmu_enable_counter(idx - 1);
+	}
 }
 
 static inline int armv8pmu_disable_counter(int idx)
@@ -545,11 +556,21 @@ static inline int armv8pmu_disable_counter(int idx)
 static inline void armv8pmu_disable_event_counter(struct perf_event *event)
 {
 	struct hw_perf_event *hwc = &event->hw;
+	struct perf_event_attr *attr = &event->attr;
 	int idx = hwc->idx;
+	u32 counter_bits = BIT(ARMV8_IDX_TO_COUNTER(idx));
 
 	if (armv8pmu_event_is_chained(event))
-		armv8pmu_disable_counter(idx - 1);
-	armv8pmu_disable_counter(idx);
+		counter_bits |= BIT(ARMV8_IDX_TO_COUNTER(idx - 1));
+
+	kvm_clr_pmu_events(counter_bits);
+
+	/* We rely on the hypervisor switch code to disable guest counters */
+	if (!kvm_pmu_counter_deferred(attr)) {
+		if (armv8pmu_event_is_chained(event))
+			armv8pmu_disable_counter(idx - 1);
+		armv8pmu_disable_counter(idx);
+	}
 }
 
 static inline int armv8pmu_enable_intens(int idx)
@@ -829,11 +850,16 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
 		if (!attr->exclude_kernel)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
 	} else {
-		if (attr->exclude_kernel)
-			config_base |= ARMV8_PMU_EXCLUDE_EL1;
-		if (!attr->exclude_hv)
+		if (!attr->exclude_hv && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
 	}
+
+	/*
+	 * Filter out !VHE kernels and guest kernels
+	 */
+	if (attr->exclude_kernel)
+		config_base |= ARMV8_PMU_EXCLUDE_EL1;
+
 	if (attr->exclude_user)
 		config_base |= ARMV8_PMU_EXCLUDE_EL0;
 
@@ -863,6 +889,9 @@ static void armv8pmu_reset(void *info)
 		armv8pmu_disable_intens(idx);
 	}
 
+	/* Clear the counters we flip at guest entry/exit */
+	kvm_clr_pmu_events(U32_MAX);
+
 	/*
 	 * Initialize & Reset PMNC. Request overflow interrupt for
 	 * 64 bit cycle counter but cheat in armv8pmu_write_counter().
Signature · 1 line
2.21.0
Acked-by on v12 4/8
#10 lore raw

Re: [PATCH v12 4/8] arm64: arm_pmu: Add !VHE support for exclude_host/exclude_guest attributes

Quoted 79 lines · Andrew Murray

Add support for the :G and :H attributes in perf by handling the exclude_host/exclude_guest event attributes.

We notify KVM of counters that we wish to be enabled or disabled on guest entry/exit and thus defer from starting or stopping events based on their event attributes.

With !VHE we switch the counters between host/guest at EL2. We are able to eliminate counters counting host events on the boundaries of guest entry/exit when using :G by filtering out EL2 for exclude_host. When using !exclude_hv there is a small blackout window at the guest entry/exit where host events are not captured.

---
 arch/arm64/kernel/perf_event.c | 43 ++++++++++++++++++++++++++++------
 1 file changed, 36 insertions(+), 7 deletions(-)
arch/arm64/kernel/perf_event.c +25 -4
index cccf4fc86d67..6bb28aaf5aea 100644
@@ -26,6 +26,7 @@
 
 #include <linux/acpi.h>
 #include <linux/clocksource.h>
+#include <linux/kvm_host.h>
 #include <linux/of.h>
 #include <linux/perf/arm_pmu.h>
 #include <linux/platform_device.h>
@@ -528,11 +529,21 @@ static inline int armv8pmu_enable_counter(int idx)
 
 static inline void armv8pmu_enable_event_counter(struct perf_event *event)
 {
+	struct perf_event_attr *attr = &event->attr;
 	int idx = event->hw.idx;
+	u32 counter_bits = BIT(ARMV8_IDX_TO_COUNTER(idx));
 
-	armv8pmu_enable_counter(idx);
 	if (armv8pmu_event_is_chained(event))
-		armv8pmu_enable_counter(idx - 1);
+		counter_bits |= BIT(ARMV8_IDX_TO_COUNTER(idx - 1));
+
+	kvm_set_pmu_events(counter_bits, attr);
+
+	/* We rely on the hypervisor switch code to enable guest counters */
+	if (!kvm_pmu_counter_deferred(attr)) {
+		armv8pmu_enable_counter(idx);
+		if (armv8pmu_event_is_chained(event))
+			armv8pmu_enable_counter(idx - 1);
+	}
 }
 
 static inline int armv8pmu_disable_counter(int idx)
@@ -545,11 +556,21 @@ static inline int armv8pmu_disable_counter(int idx)
 static inline void armv8pmu_disable_event_counter(struct perf_event *event)
 {
 	struct hw_perf_event *hwc = &event->hw;
+	struct perf_event_attr *attr = &event->attr;
 	int idx = hwc->idx;
+	u32 counter_bits = BIT(ARMV8_IDX_TO_COUNTER(idx));
 
 	if (armv8pmu_event_is_chained(event))
-		armv8pmu_disable_counter(idx - 1);
-	armv8pmu_disable_counter(idx);
+		counter_bits |= BIT(ARMV8_IDX_TO_COUNTER(idx - 1));
+
+	kvm_clr_pmu_events(counter_bits);
+
+	/* We rely on the hypervisor switch code to disable guest counters */
+	if (!kvm_pmu_counter_deferred(attr)) {
+		if (armv8pmu_event_is_chained(event))
+			armv8pmu_disable_counter(idx - 1);
+		armv8pmu_disable_counter(idx);
+	}
 }
 
 static inline int armv8pmu_enable_intens(int idx)

arch/arm64/kernel/perf_event.c static int armv8pmu_set_event_filter(struct hw_perf_event *event, · line 854

Quoted 8 lines
arch/arm64/kernel/perf_event.c +1 -3
@@ -829,11 +850,16 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
 		if (!attr->exclude_kernel)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
 	} else {
-		if (attr->exclude_kernel)
-			config_base |= ARMV8_PMU_EXCLUDE_EL1;
-		if (!attr->exclude_hv)
+		if (!attr->exclude_hv && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;

FWIW, that doesn't align with my personal view of what the host is, but I'm not going to argue if Marc and Christoffer are happy with it.

Anyway, thanks for persevering with this:

Will

#11 lore raw

Re: [PATCH v12 4/8] arm64: arm_pmu: Add !VHE support for exclude_host/exclude_guest attributes

Quoted 78 lines · Will Deacon
Quoted 77 lines · Andrew Murray

Add support for the :G and :H attributes in perf by handling the exclude_host/exclude_guest event attributes.

We notify KVM of counters that we wish to be enabled or disabled on guest entry/exit and thus defer from starting or stopping events based on their event attributes.

With !VHE we switch the counters between host/guest at EL2. We are able to eliminate counters counting host events on the boundaries of guest entry/exit when using :G by filtering out EL2 for exclude_host. When using !exclude_hv there is a small blackout window at the guest entry/exit where host events are not captured.

---
 arch/arm64/kernel/perf_event.c | 43 ++++++++++++++++++++++++++++------
 1 file changed, 36 insertions(+), 7 deletions(-)
arch/arm64/kernel/perf_event.c +25 -4
index cccf4fc86d67..6bb28aaf5aea 100644
@@ -26,6 +26,7 @@
 
 #include <linux/acpi.h>
 #include <linux/clocksource.h>
+#include <linux/kvm_host.h>
 #include <linux/of.h>
 #include <linux/perf/arm_pmu.h>
 #include <linux/platform_device.h>
@@ -528,11 +529,21 @@ static inline int armv8pmu_enable_counter(int idx)
 
 static inline void armv8pmu_enable_event_counter(struct perf_event *event)
 {
+	struct perf_event_attr *attr = &event->attr;
 	int idx = event->hw.idx;
+	u32 counter_bits = BIT(ARMV8_IDX_TO_COUNTER(idx));
 
-	armv8pmu_enable_counter(idx);
 	if (armv8pmu_event_is_chained(event))
-		armv8pmu_enable_counter(idx - 1);
+		counter_bits |= BIT(ARMV8_IDX_TO_COUNTER(idx - 1));
+
+	kvm_set_pmu_events(counter_bits, attr);
+
+	/* We rely on the hypervisor switch code to enable guest counters */
+	if (!kvm_pmu_counter_deferred(attr)) {
+		armv8pmu_enable_counter(idx);
+		if (armv8pmu_event_is_chained(event))
+			armv8pmu_enable_counter(idx - 1);
+	}
 }
 
 static inline int armv8pmu_disable_counter(int idx)
@@ -545,11 +556,21 @@ static inline int armv8pmu_disable_counter(int idx)
 static inline void armv8pmu_disable_event_counter(struct perf_event *event)
 {
 	struct hw_perf_event *hwc = &event->hw;
+	struct perf_event_attr *attr = &event->attr;
 	int idx = hwc->idx;
+	u32 counter_bits = BIT(ARMV8_IDX_TO_COUNTER(idx));
 
 	if (armv8pmu_event_is_chained(event))
-		armv8pmu_disable_counter(idx - 1);
-	armv8pmu_disable_counter(idx);
+		counter_bits |= BIT(ARMV8_IDX_TO_COUNTER(idx - 1));
+
+	kvm_clr_pmu_events(counter_bits);
+
+	/* We rely on the hypervisor switch code to disable guest counters */
+	if (!kvm_pmu_counter_deferred(attr)) {
+		if (armv8pmu_event_is_chained(event))
+			armv8pmu_disable_counter(idx - 1);
+		armv8pmu_disable_counter(idx);
+	}
 }
 
Quoted 13 lines

arch/arm64/kernel/perf_event.c static int armv8pmu_set_event_filter(struct hw_perf_event *event, · line 854

Quoted 10 lines
arch/arm64/kernel/perf_event.c +1 -3
 static inline int armv8pmu_enable_intens(int idx)
@@ -829,11 +850,16 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
 		if (!attr->exclude_kernel)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
 	} else {
-		if (attr->exclude_kernel)
-			config_base |= ARMV8_PMU_EXCLUDE_EL1;
-		if (!attr->exclude_hv)
+		if (!attr->exclude_hv && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;

FWIW, that doesn't align with my personal view of what the host is, but I'm not going to argue if Marc and Christoffer are happy with it.

Ideally armv8pmu_set_event_filter needs no knowledge of exclude_guest and exclude_host as we leave the config_base alone and instead rely on turning the counters on/off at guest entry/exit...

The addition of "&& !attr->exclude_host" here (!VHE) and likewise for VHE in "arm64: KVM: Enable VHE support for :G/:H perf event modifiers" (in case you missed it) is to eliminate counters counting host events on the boundaries of guest entry/exit when counting for guest_only (:G). Thus this is really an optimisation for more accurate counting.

Consider the case where a user wants to record guest events only - we switch over to the guest counters at EL2 on world-switch (both VHE and !VHE host) - now there is a small period of time between when we start the guest counters and when we actually jump into the guest. This period of time in the host shouldn't be counted - therefore so long as exclude_host is set we can filter EL2 to remove these unwanted counts.

Therefore the above hunks likely won't describe what our view of the host is.

Quoted 2 lines

Anyway, thanks for persevering with this:

It's fun even though it hurts my head every time each time I come back to this.

Quoted 2 lines

Thanks,

Andrew Murray

Quoted 2 lines

Will

#12 lore raw

PATCH v12 5/8 arm64: KVM: Enable !VHE support for :G/:H perf event modifiers

In reply to Andrew Murray

Enable/disable event counters as appropriate when entering and exiting the guest to enable support for guest or host only event counting.

For both VHE and non-VHE we switch the counters between host/guest at EL2.

The PMU may be on when we change which counters are enabled however we avoid adding an isb as we instead rely on existing context synchronisation events: the eret to enter the guest (__guest_enter) and eret in kvm_call_hyp for __kvm_vcpu_run_nvhe on returning.

---
 arch/arm64/include/asm/kvm_host.h |  3 +++
 arch/arm64/kvm/hyp/switch.c       |  6 +++++
 arch/arm64/kvm/pmu.c              | 40 +++++++++++++++++++++++++++++++
 3 files changed, 49 insertions(+)
arch/arm64/include/asm/kvm_host.h +3 -0
index 213465196e6b..a3bfb75f0be9 100644
@@ -539,6 +539,9 @@ static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
 
 void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr);
 void kvm_clr_pmu_events(u32 clr);
+
+void __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt);
+bool __pmu_switch_to_guest(struct kvm_cpu_context *host_ctxt);
 #else
 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
 static inline void kvm_clr_pmu_events(u32 clr) {}
arch/arm64/kvm/hyp/switch.c +6 -0
index 3563fe655cd5..d1dc40e53f22 100644
@@ -524,6 +524,7 @@ int __hyp_text __kvm_vcpu_run_nvhe(struct kvm_vcpu *vcpu)
 {
 	struct kvm_cpu_context *host_ctxt;
 	struct kvm_cpu_context *guest_ctxt;
+	bool pmu_switch_needed;
 	u64 exit_code;
 
 	/*
@@ -543,6 +544,8 @@ int __hyp_text __kvm_vcpu_run_nvhe(struct kvm_vcpu *vcpu)
 	host_ctxt->__hyp_running_vcpu = vcpu;
 	guest_ctxt = &vcpu->arch.ctxt;
 
+	pmu_switch_needed = __pmu_switch_to_guest(host_ctxt);
+
 	__sysreg_save_state_nvhe(host_ctxt);
 
 	__activate_vm(kern_hyp_va(vcpu->kvm));
@@ -589,6 +592,9 @@ int __hyp_text __kvm_vcpu_run_nvhe(struct kvm_vcpu *vcpu)
 	 */
 	__debug_switch_to_host(vcpu);
 
+	if (pmu_switch_needed)
+		__pmu_switch_to_host(host_ctxt);
+
 	/* Returning to host will clear PSR.I, remask PMR if needed */
 	if (system_uses_irq_prio_masking())
 		gic_write_pmr(GIC_PRIO_IRQOFF);
arch/arm64/kvm/pmu.c +40 -0
index 5414b134f99a..48949f0cdba0 100644
@@ -5,6 +5,7 @@
  */
 #include <linux/kvm_host.h>
 #include <linux/perf_event.h>
+#include <asm/kvm_hyp.h>
 
 /*
  * Given the exclude_{host,guest} attributes, determine if we are going
@@ -43,3 +44,42 @@ void kvm_clr_pmu_events(u32 clr)
 	ctx->pmu_events.events_host &= ~clr;
 	ctx->pmu_events.events_guest &= ~clr;
 }
+
+/**
+ * Disable host events, enable guest events
+ */
+bool __hyp_text __pmu_switch_to_guest(struct kvm_cpu_context *host_ctxt)
+{
+	struct kvm_host_data *host;
+	struct kvm_pmu_events *pmu;
+
+	host = container_of(host_ctxt, struct kvm_host_data, host_ctxt);
+	pmu = &host->pmu_events;
+
+	if (pmu->events_host)
+		write_sysreg(pmu->events_host, pmcntenclr_el0);
+
+	if (pmu->events_guest)
+		write_sysreg(pmu->events_guest, pmcntenset_el0);
+
+	return (pmu->events_host || pmu->events_guest);
+}
+
+/**
+ * Disable guest events, enable host events
+ */
+void __hyp_text __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt)
+{
+	struct kvm_host_data *host;
+	struct kvm_pmu_events *pmu;
+
+	host = container_of(host_ctxt, struct kvm_host_data, host_ctxt);
+	pmu = &host->pmu_events;
+
+	if (pmu->events_guest)
+		write_sysreg(pmu->events_guest, pmcntenclr_el0);
+
+	if (pmu->events_host)
+		write_sysreg(pmu->events_host, pmcntenset_el0);
+}
+
Signature · 1 line
2.21.0
#13 lore raw

PATCH v12 6/8 arm64: KVM: Enable VHE support for :G/:H perf event modifiers

In reply to Andrew Murray

With VHE different exception levels are used between the host (EL2) and guest (EL1) with a shared exception level for userpace (EL0). We can take advantage of this and use the PMU's exception level filtering to avoid enabling/disabling counters in the world-switch code. Instead we just modify the counter type to include or exclude EL0 at vcpu_{load,put} time.

We also ensure that trapped PMU system register writes do not re-enable EL0 when reconfiguring the backing perf events.

This approach completely avoids blackout windows seen with !VHE.

---
 arch/arm/include/asm/kvm_host.h   |  3 ++
 arch/arm64/include/asm/kvm_host.h |  5 +-
 arch/arm64/kernel/perf_event.c    |  6 ++-
 arch/arm64/kvm/pmu.c              | 87 ++++++++++++++++++++++++++++++-
 arch/arm64/kvm/sys_regs.c         |  3 ++
 virt/kvm/arm/arm.c                |  2 +
 6 files changed, 102 insertions(+), 4 deletions(-)
arch/arm/include/asm/kvm_host.h +3 -0
index 427c28be6452..481411295b3b 100644
@@ -365,6 +365,9 @@ static inline void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu) {}
 
+static inline void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu) {}
+static inline void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu) {}
+
 static inline void kvm_arm_vhe_guest_enter(void) {}
 static inline void kvm_arm_vhe_guest_exit(void) {}
 
arch/arm64/include/asm/kvm_host.h +4 -1
index a3bfb75f0be9..4f290dad3a48 100644
@@ -528,7 +528,7 @@ void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
 
 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
 {
-	return attr->exclude_host;
+	return (!has_vhe() && attr->exclude_host);
 }
 
 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
@@ -542,6 +542,9 @@ void kvm_clr_pmu_events(u32 clr);
 
 void __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt);
 bool __pmu_switch_to_guest(struct kvm_cpu_context *host_ctxt);
+
+void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
+void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
 #else
 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
 static inline void kvm_clr_pmu_events(u32 clr) {}
arch/arm64/kernel/perf_event.c +5 -1
index 6bb28aaf5aea..314b1adedf06 100644
@@ -847,8 +847,12 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
 	 * with other architectures (x86 and Power).
 	 */
 	if (is_kernel_in_hyp_mode()) {
-		if (!attr->exclude_kernel)
+		if (!attr->exclude_kernel && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
+		if (attr->exclude_guest)
+			config_base |= ARMV8_PMU_EXCLUDE_EL1;
+		if (attr->exclude_host)
+			config_base |= ARMV8_PMU_EXCLUDE_EL0;
 	} else {
 		if (!attr->exclude_hv && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
arch/arm64/kvm/pmu.c +85 -2

99 lines not on this page · raw message

arch/arm64/kvm/sys_regs.c +3 -0
index 539feecda5b8..c7fa47ad2387 100644
@@ -695,6 +695,7 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
 		val |= p->regval & ARMV8_PMU_PMCR_MASK;
 		__vcpu_sys_reg(vcpu, PMCR_EL0) = val;
 		kvm_pmu_handle_pmcr(vcpu, val);
+		kvm_vcpu_pmu_restore_guest(vcpu);
 	} else {
 		/* PMCR.P & PMCR.C are RAZ */
 		val = __vcpu_sys_reg(vcpu, PMCR_EL0)
@@ -850,6 +851,7 @@ static bool access_pmu_evtyper(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
 	if (p->is_write) {
 		kvm_pmu_set_counter_event_type(vcpu, p->regval, idx);
 		__vcpu_sys_reg(vcpu, reg) = p->regval & ARMV8_PMU_EVTYPE_MASK;
+		kvm_vcpu_pmu_restore_guest(vcpu);
 	} else {
 		p->regval = __vcpu_sys_reg(vcpu, reg) & ARMV8_PMU_EVTYPE_MASK;
 	}
@@ -875,6 +877,7 @@ static bool access_pmcnten(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
 			/* accessing PMCNTENSET_EL0 */
 			__vcpu_sys_reg(vcpu, PMCNTENSET_EL0) |= val;
 			kvm_pmu_enable_counter(vcpu, val);
+			kvm_vcpu_pmu_restore_guest(vcpu);
 		} else {
 			/* accessing PMCNTENCLR_EL0 */
 			__vcpu_sys_reg(vcpu, PMCNTENSET_EL0) &= ~val;
virt/kvm/arm/arm.c +2 -0
index 6bcda52c0ea6..d805bb470dda 100644
@@ -382,6 +382,7 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 	kvm_timer_vcpu_load(vcpu);
 	kvm_vcpu_load_sysregs(vcpu);
 	kvm_arch_vcpu_load_fp(vcpu);
+	kvm_vcpu_pmu_restore_guest(vcpu);
 
 	if (single_task_running())
 		vcpu_clear_wfe_traps(vcpu);
@@ -395,6 +396,7 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 	kvm_vcpu_put_sysregs(vcpu);
 	kvm_timer_vcpu_put(vcpu);
 	kvm_vgic_put(vcpu);
+	kvm_vcpu_pmu_restore_host(vcpu);
 
 	vcpu->cpu = -1;
 
Signature · 1 line
2.21.0
Acked-by on v12 6/8
#14 lore raw

Re: [PATCH v12 6/8] arm64: KVM: Enable VHE support for :G/:H perf event modifiers

Quoted 67 lines · Andrew Murray

With VHE different exception levels are used between the host (EL2) and guest (EL1) with a shared exception level for userpace (EL0). We can take advantage of this and use the PMU's exception level filtering to avoid enabling/disabling counters in the world-switch code. Instead we just modify the counter type to include or exclude EL0 at vcpu_{load,put} time.

We also ensure that trapped PMU system register writes do not re-enable EL0 when reconfiguring the backing perf events.

This approach completely avoids blackout windows seen with !VHE.

---
 arch/arm/include/asm/kvm_host.h   |  3 ++
 arch/arm64/include/asm/kvm_host.h |  5 +-
 arch/arm64/kernel/perf_event.c    |  6 ++-
 arch/arm64/kvm/pmu.c              | 87 ++++++++++++++++++++++++++++++-
 arch/arm64/kvm/sys_regs.c         |  3 ++
 virt/kvm/arm/arm.c                |  2 +
 6 files changed, 102 insertions(+), 4 deletions(-)
arch/arm/include/asm/kvm_host.h +3 -0
index 427c28be6452..481411295b3b 100644
@@ -365,6 +365,9 @@ static inline void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu) {}
 
+static inline void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu) {}
+static inline void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu) {}
+
 static inline void kvm_arm_vhe_guest_enter(void) {}
 static inline void kvm_arm_vhe_guest_exit(void) {}
 
arch/arm64/include/asm/kvm_host.h +4 -1
index a3bfb75f0be9..4f290dad3a48 100644
@@ -528,7 +528,7 @@ void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
 
 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
 {
-	return attr->exclude_host;
+	return (!has_vhe() && attr->exclude_host);
 }
 
 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
@@ -542,6 +542,9 @@ void kvm_clr_pmu_events(u32 clr);
 
 void __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt);
 bool __pmu_switch_to_guest(struct kvm_cpu_context *host_ctxt);
+
+void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
+void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
 #else
 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
 static inline void kvm_clr_pmu_events(u32 clr) {}
arch/arm64/kernel/perf_event.c +0 -0
index 6bb28aaf5aea..314b1adedf06 100644
@@ -847,8 +847,12 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
 	 * with other architectures (x86 and Power).
 	 */
 	if (is_kernel_in_hyp_mode()) {

arch/arm64/kernel/perf_event.c static int armv8pmu_set_event_filter(struct hw_perf_event *event, · line 858

Quoted 10 lines
arch/arm64/kernel/perf_event.c +5 -1
-		if (!attr->exclude_kernel)
+		if (!attr->exclude_kernel && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
+		if (attr->exclude_guest)
+			config_base |= ARMV8_PMU_EXCLUDE_EL1;
+		if (attr->exclude_host)
+			config_base |= ARMV8_PMU_EXCLUDE_EL0;
 	} else {
 		if (!attr->exclude_hv && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;

I still don't really like these semantics, but it's consistent and you're documenting it so:

Will

#15 lore raw

Re: [PATCH v12 6/8] arm64: KVM: Enable VHE support for :G/:H perf event modifiers

Quoted 82 lines · Will Deacon
Quoted 67 lines · Andrew Murray

With VHE different exception levels are used between the host (EL2) and guest (EL1) with a shared exception level for userpace (EL0). We can take advantage of this and use the PMU's exception level filtering to avoid enabling/disabling counters in the world-switch code. Instead we just modify the counter type to include or exclude EL0 at vcpu_{load,put} time.

We also ensure that trapped PMU system register writes do not re-enable EL0 when reconfiguring the backing perf events.

This approach completely avoids blackout windows seen with !VHE.

---
 arch/arm/include/asm/kvm_host.h   |  3 ++
 arch/arm64/include/asm/kvm_host.h |  5 +-
 arch/arm64/kernel/perf_event.c    |  6 ++-
 arch/arm64/kvm/pmu.c              | 87 ++++++++++++++++++++++++++++++-
 arch/arm64/kvm/sys_regs.c         |  3 ++
 virt/kvm/arm/arm.c                |  2 +
 6 files changed, 102 insertions(+), 4 deletions(-)
arch/arm/include/asm/kvm_host.h +3 -0
index 427c28be6452..481411295b3b 100644
@@ -365,6 +365,9 @@ static inline void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu) {}
 static inline void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu) {}
 
+static inline void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu) {}
+static inline void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu) {}
+
 static inline void kvm_arm_vhe_guest_enter(void) {}
 static inline void kvm_arm_vhe_guest_exit(void) {}
 
arch/arm64/include/asm/kvm_host.h +4 -1
index a3bfb75f0be9..4f290dad3a48 100644
@@ -528,7 +528,7 @@ void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu);
 
 static inline bool kvm_pmu_counter_deferred(struct perf_event_attr *attr)
 {
-	return attr->exclude_host;
+	return (!has_vhe() && attr->exclude_host);
 }
 
 #ifdef CONFIG_KVM /* Avoid conflicts with core headers if CONFIG_KVM=n */
@@ -542,6 +542,9 @@ void kvm_clr_pmu_events(u32 clr);
 
 void __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt);
 bool __pmu_switch_to_guest(struct kvm_cpu_context *host_ctxt);
+
+void kvm_vcpu_pmu_restore_guest(struct kvm_vcpu *vcpu);
+void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu);
 #else
 static inline void kvm_set_pmu_events(u32 set, struct perf_event_attr *attr) {}
 static inline void kvm_clr_pmu_events(u32 clr) {}
arch/arm64/kernel/perf_event.c +0 -0
index 6bb28aaf5aea..314b1adedf06 100644
@@ -847,8 +847,12 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
 	 * with other architectures (x86 and Power).
 	 */
 	if (is_kernel_in_hyp_mode()) {

arch/arm64/kernel/perf_event.c static int armv8pmu_set_event_filter(struct hw_perf_event *event, · line 858

Quoted 10 lines
arch/arm64/kernel/perf_event.c +5 -1
-		if (!attr->exclude_kernel)
+		if (!attr->exclude_kernel && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;
+		if (attr->exclude_guest)
+			config_base |= ARMV8_PMU_EXCLUDE_EL1;
+		if (attr->exclude_host)
+			config_base |= ARMV8_PMU_EXCLUDE_EL0;
 	} else {
 		if (!attr->exclude_hv && !attr->exclude_host)
 			config_base |= ARMV8_PMU_INCLUDE_EL2;

I still don't really like these semantics, but it's consistent and you're documenting it so:

Quoted 1 line

Much appreciated.

Thanks,

Andrew Murray

Quoted 2 lines

Will

#16 lore raw

PATCH v12 7/8 arm64: KVM: avoid isb's by using direct pmxevtyper sysreg

In reply to Andrew Murray

Upon entering or exiting a guest we may modify multiple PMU counters to enable of disable EL0 filtering. We presently do this via the indirect PMXEVTYPER_EL0 system register (where the counter we modify is selected by PMSELR). With this approach it is necessary to order the writes via isb instructions such that we select the correct counter before modifying it.

Let's avoid potentially expensive instruction barriers by using the direct PMEVTYPER<n>_EL0 registers instead.

As the change to counter type relates only to EL0 filtering we can rely on the implicit instruction barrier which occurs when we transition from EL2 to EL1 on entering the guest. On returning to userspace we can, at the latest, rely on the implicit barrier between EL2 and EL0. We can also depend on the explicit isb in armv8pmu_select_counter to order our write against any other kernel changes by the PMU driver to the type register as a result of preemption.

---
 arch/arm64/kvm/pmu.c | 84 ++++++++++++++++++++++++++++++++++++++------
 1 file changed, 74 insertions(+), 10 deletions(-)
arch/arm64/kvm/pmu.c +74 -10
index 3407a529e116..4d55193ccc71 100644
@@ -91,6 +91,74 @@ void __hyp_text __pmu_switch_to_host(struct kvm_cpu_context *host_ctxt)
 		write_sysreg(pmu->events_host, pmcntenset_el0);
 }
 
+#define PMEVTYPER_READ_CASE(idx)				\
+	case idx:						\
+		return read_sysreg(pmevtyper##idx##_el0)
+
+#define PMEVTYPER_WRITE_CASE(idx)				\
+	case idx:						\
+		write_sysreg(val, pmevtyper##idx##_el0);	\
+		break
+
+#define PMEVTYPER_CASES(readwrite)				\
+	PMEVTYPER_##readwrite##_CASE(0);			\
+	PMEVTYPER_##readwrite##_CASE(1);			\
+	PMEVTYPER_##readwrite##_CASE(2);			\
+	PMEVTYPER_##readwrite##_CASE(3);			\
+	PMEVTYPER_##readwrite##_CASE(4);			\
+	PMEVTYPER_##readwrite##_CASE(5);			\
+	PMEVTYPER_##readwrite##_CASE(6);			\
+	PMEVTYPER_##readwrite##_CASE(7);			\
+	PMEVTYPER_##readwrite##_CASE(8);			\
+	PMEVTYPER_##readwrite##_CASE(9);			\
+	PMEVTYPER_##readwrite##_CASE(10);			\
+	PMEVTYPER_##readwrite##_CASE(11);			\
+	PMEVTYPER_##readwrite##_CASE(12);			\
+	PMEVTYPER_##readwrite##_CASE(13);			\
+	PMEVTYPER_##readwrite##_CASE(14);			\
+	PMEVTYPER_##readwrite##_CASE(15);			\
+	PMEVTYPER_##readwrite##_CASE(16);			\
+	PMEVTYPER_##readwrite##_CASE(17);			\
+	PMEVTYPER_##readwrite##_CASE(18);			\
+	PMEVTYPER_##readwrite##_CASE(19);			\
+	PMEVTYPER_##readwrite##_CASE(20);			\
+	PMEVTYPER_##readwrite##_CASE(21);			\
+	PMEVTYPER_##readwrite##_CASE(22);			\
+	PMEVTYPER_##readwrite##_CASE(23);			\
+	PMEVTYPER_##readwrite##_CASE(24);			\
+	PMEVTYPER_##readwrite##_CASE(25);			\
+	PMEVTYPER_##readwrite##_CASE(26);			\
+	PMEVTYPER_##readwrite##_CASE(27);			\
+	PMEVTYPER_##readwrite##_CASE(28);			\
+	PMEVTYPER_##readwrite##_CASE(29);			\
+	PMEVTYPER_##readwrite##_CASE(30)
+
+/*
+ * Read a value direct from PMEVTYPER<idx>
+ */
+static u64 kvm_vcpu_pmu_read_evtype_direct(int idx)
+{
+	switch (idx) {
+	PMEVTYPER_CASES(READ);
+	default:
+		WARN_ON(1);
+	}
+
+	return 0;
+}
+
+/*
+ * Write a value direct to PMEVTYPER<idx>
+ */
+static void kvm_vcpu_pmu_write_evtype_direct(int idx, u32 val)
+{
+	switch (idx) {
+	PMEVTYPER_CASES(WRITE);
+	default:
+		WARN_ON(1);
+	}
+}
+
 /*
  * Modify ARMv8 PMU events to include EL0 counting
  */
@@ -100,11 +168,9 @@ static void kvm_vcpu_pmu_enable_el0(unsigned long events)
 	u32 counter;
 
 	for_each_set_bit(counter, &events, 32) {
-		write_sysreg(counter, pmselr_el0);
-		isb();
-		typer = read_sysreg(pmxevtyper_el0) & ~ARMV8_PMU_EXCLUDE_EL0;
-		write_sysreg(typer, pmxevtyper_el0);
-		isb();
+		typer = kvm_vcpu_pmu_read_evtype_direct(counter);
+		typer &= ~ARMV8_PMU_EXCLUDE_EL0;
+		kvm_vcpu_pmu_write_evtype_direct(counter, typer);
 	}
 }
 
@@ -117,11 +183,9 @@ static void kvm_vcpu_pmu_disable_el0(unsigned long events)
 	u32 counter;
 
 	for_each_set_bit(counter, &events, 32) {
-		write_sysreg(counter, pmselr_el0);
-		isb();
-		typer = read_sysreg(pmxevtyper_el0) | ARMV8_PMU_EXCLUDE_EL0;
-		write_sysreg(typer, pmxevtyper_el0);
-		isb();
+		typer = kvm_vcpu_pmu_read_evtype_direct(counter);
+		typer |= ARMV8_PMU_EXCLUDE_EL0;
+		kvm_vcpu_pmu_write_evtype_direct(counter, typer);
 	}
 }
 
Signature · 1 line
2.21.0
#17 lore raw

PATCH v12 8/8 arm64: docs: document perf event attributes

In reply to Andrew Murray

The interaction between the exclude_{host,guest} flags, exclude_{user,kernel,hv} flags and presence of VHE can result in different exception levels being filtered by the ARMv8 PMU. As this can be confusing let's document how they work on arm64.

---
 Documentation/arm64/perf.txt | 74 ++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)
 create mode 100644 Documentation/arm64/perf.txt
Documentation/arm64/perf.txt new file +74 -0
new file mode 100644
index 000000000000..604446c1f720
@@ -0,0 +1,74 @@
+Perf Event Attributes
+=====================
+
+Author: Andrew Murray <[email protected]>
+Date: 2019-03-06
+
+exclude_user
+------------
+
+This attribute excludes userspace.
+
+Userspace always runs at EL0 and thus this attribute will exclude EL0.
+
+
+exclude_kernel
+--------------
+
+This attribute excludes the kernel.
+
+The kernel runs at EL2 with VHE and EL1 without. Guest kernels always run
+at EL1.
+
+This attribute will exclude EL1 and additionally EL2 on a VHE system.
+
+
+exclude_hv
+----------
+
+This attribute excludes the hypervisor, we ignore this flag on a VHE system
+as we consider the host kernel to be the hypervisor.
+
+On a non-VHE system we consider the hypervisor to be any code that runs at
+EL2 which is predominantly used for guest/host transitions.
+
+This attribute will exclude EL2 on a non-VHE system.
+
+
+exclude_host / exclude_guest
+----------------------------
+
+This attribute excludes the KVM host.
+
+The KVM host may run at EL0 (userspace), EL1 (non-VHE kernel) and EL2 (VHE
+kernel or non-VHE hypervisor).
+
+The KVM guest may run at EL0 (userspace) and EL1 (kernel).
+
+Due to the overlapping exception levels between host and guests we cannot
+exclusively rely on the PMU's hardware exception filtering - therefore we
+must enable/disable counting on the entry and exit to the guest. This is
+performed differently on VHE and non-VHE systems.
+
+For non-VHE systems we exclude EL2 for exclude_host - upon entering and
+exiting the guest we disable/enable the event as appropriate based on the
+exclude_host and exclude_guest attributes.
+
+For VHE systems we exclude EL1 for exclude_guest and exclude both EL0,EL2
+for exclude_host. Upon entering and exiting the guest we modify the event
+to include/exclude EL0 as appropriate based on the exclude_host and
+exclude_guest attributes.
+
+
+Accuracy
+--------
+
+On non-VHE systems we enable/disable counters on the entry/exit of
+host/guest transition at EL2 - however there is a period of time between
+enabling/disabling the counters and entering/exiting the guest. We are
+able to eliminate counters counting host events on the boundaries of guest
+entry/exit when counting guest events by filtering out EL2 for
+exclude_host. However when using !exclude_hv there is a small blackout
+window at the guest entry/exit where host events are not captured.
+
+On VHE systems there are no blackout windows.
Signature · 1 line
2.21.0
#18 lore raw

Re: [PATCH v12 8/8] arm64: docs: document perf event attributes

Quoted 30 lines · Andrew Murray

The interaction between the exclude_{host,guest} flags, exclude_{user,kernel,hv} flags and presence of VHE can result in different exception levels being filtered by the ARMv8 PMU. As this can be confusing let's document how they work on arm64.

---
 Documentation/arm64/perf.txt | 74 ++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)
 create mode 100644 Documentation/arm64/perf.txt
Documentation/arm64/perf.txt new file +13 -0
new file mode 100644
index 000000000000..604446c1f720
@@ -0,0 +1,74 @@
+Perf Event Attributes
+=====================
+
+Author: Andrew Murray <[email protected]>
+Date: 2019-03-06
+
+exclude_user
+------------
+
+This attribute excludes userspace.
+
+Userspace always runs at EL0 and thus this attribute will exclude EL0.
+

Documentation/arm64/perf.txt line 23

Quoted 10 lines
Documentation/arm64/perf.txt new file +10 -0
+
+exclude_kernel
+--------------
+
+This attribute excludes the kernel.
+
+The kernel runs at EL2 with VHE and EL1 without. Guest kernels always run
+at EL1.
+
+This attribute will exclude EL1 and additionally EL2 on a VHE system.

I find this last sentence a bit confusing, because it can be read to imply that if you don't set exclude_kernel and you're in a guest on a VHE system, then you can profile EL2.

Documentation/arm64/perf.txt line 30

Quoted 5 lines
Documentation/arm64/perf.txt new file +5 -0
+exclude_hv
+----------
+
+This attribute excludes the hypervisor, we ignore this flag on a VHE system
+as we consider the host kernel to be the hypervisor.

Similar comment as the above: I don't think this makes sense when you look at things from the guest perspective.

Documentation/arm64/perf.txt line 41

Quoted 10 lines
Documentation/arm64/perf.txt new file +10 -0
+On a non-VHE system we consider the hypervisor to be any code that runs at
+EL2 which is predominantly used for guest/host transitions.
+
+This attribute will exclude EL2 on a non-VHE system.
+
+
+exclude_host / exclude_guest
+----------------------------
+
+This attribute excludes the KVM host.

But there are two attributes...

Will

#19 lore raw

Re: [PATCH v12 8/8] arm64: docs: document perf event attributes

Quoted 37 lines · Will Deacon
Quoted 36 lines · Andrew Murray

The interaction between the exclude_{host,guest} flags, exclude_{user,kernel,hv} flags and presence of VHE can result in different exception levels being filtered by the ARMv8 PMU. As this can be confusing let's document how they work on arm64.

---
 Documentation/arm64/perf.txt | 74 ++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)
 create mode 100644 Documentation/arm64/perf.txt
Documentation/arm64/perf.txt new file +19 -0
new file mode 100644
index 000000000000..604446c1f720
@@ -0,0 +1,74 @@
+Perf Event Attributes
+=====================
+
+Author: Andrew Murray <[email protected]>
+Date: 2019-03-06
+
+exclude_user
+------------
+
+This attribute excludes userspace.
+
+Userspace always runs at EL0 and thus this attribute will exclude EL0.
+
+
+exclude_kernel
+--------------
+
+This attribute excludes the kernel.
+
Quoted 8 lines

Documentation/arm64/perf.txt line 23

Quoted 4 lines
Documentation/arm64/perf.txt new file +4 -0
+The kernel runs at EL2 with VHE and EL1 without. Guest kernels always run
+at EL1.
+
+This attribute will exclude EL1 and additionally EL2 on a VHE system.

I find this last sentence a bit confusing, because it can be read to imply that if you don't set exclude_kernel and you're in a guest on a VHE system, then you can profile EL2.

Yes this could be misleading.

However from the perspective of the guest, when exclude_kernel is not set we do indeed allow the guest to program it's PMU with ARMV8_PMU_INCLUDE_EL2 - and thus the statement above is correct in terms of what the kernel believes it is doing.

I think these statements are less confusing if we treat the exception levels as those 'detected' by the running context (e.g. consider the impact of nested virt here) - and we if ignore what the hypervisor (KVM) does outside (e.g. stops counting upon switching between guest/host, translating PMU filters in kvm_pmu_set_counter_event_type etc, etc). This then makes this document useful for those wishing to change this logic (which is the intent) rather than those trying to understand how we filter for EL levels as seen bare-metal.

With regards to the example you gave (exclude_kernel, EL2) - yes we want the kernel to believe it can count EL2 - because one day we may want to update KVM to allow the guest to count it's hypervisor overhead (e.g. host kernel time associated with the guest).

I could write some preface that describes this outlook. Alternatively I could just spell out what happens on a guest, e.g.

"For the host this attribute will exclude EL1 and additionally EL2 on a VHE system.

For the guest this attribute will exclude EL1."

Though I'm less comfortable with this, as the last statement "For the guest this attribute will exclude EL1." describes the product of both kvm_pmu_set_counter_event_type and armv8pmu_set_event_filter which is confusing to work out and also makes an assumption that we don't have nested virt (true for now at least) and also reasons about bare-metal EL levels which probably aren't that useful for someone changing this logic or understanding what the flags do for there performance analysis.

Do you have a preference for how this is improved?

Quoted 10 lines

Documentation/arm64/perf.txt line 30

Quoted 6 lines
Documentation/arm64/perf.txt new file +5 -0
+exclude_hv
+----------
+
+This attribute excludes the hypervisor, we ignore this flag on a VHE system
+as we consider the host kernel to be the hypervisor.

Similar comment as the above: I don't think this makes sense when you look at things from the guest perspective.

Quoted 12 lines

Documentation/arm64/perf.txt line 41

Quoted 10 lines
Documentation/arm64/perf.txt new file +10 -0
+On a non-VHE system we consider the hypervisor to be any code that runs at
+EL2 which is predominantly used for guest/host transitions.
+
+This attribute will exclude EL2 on a non-VHE system.
+
+
+exclude_host / exclude_guest
+----------------------------
+
+This attribute excludes the KVM host.

But there are two attributes...

Oh, I'll have to update this.

Thanks for the review,

Andrew Murray

Quoted 2 lines

Will

#20 lore raw

Re: [PATCH v12 8/8] arm64: docs: document perf event attributes

Quoted 31 lines · Andrew Murray
Quoted 14 lines · Will Deacon

Documentation/arm64/perf.txt line 23

Quoted 9 lines · Andrew Murray
Documentation/arm64/perf.txt new file +9 -0
+exclude_kernel
+--------------
+
+This attribute excludes the kernel.
+
+The kernel runs at EL2 with VHE and EL1 without. Guest kernels always run
+at EL1.
+
+This attribute will exclude EL1 and additionally EL2 on a VHE system.

I find this last sentence a bit confusing, because it can be read to imply that if you don't set exclude_kernel and you're in a guest on a VHE system, then you can profile EL2.

Yes this could be misleading.

However from the perspective of the guest, when exclude_kernel is not set we do indeed allow the guest to program it's PMU with ARMV8_PMU_INCLUDE_EL2 - and thus the statement above is correct in terms of what the kernel believes it is doing.

I think these statements are less confusing if we treat the exception levels as those 'detected' by the running context (e.g. consider the impact of nested virt here) - and we if ignore what the hypervisor (KVM) does outside (e.g. stops counting upon switching between guest/host, translating PMU filters in kvm_pmu_set_counter_event_type etc, etc). This then makes this document useful for those wishing to change this logic (which is the intent) rather than those trying to understand how we filter for EL levels as seen bare-metal.

Quoted 4 lines

With regards to the example you gave (exclude_kernel, EL2) - yes we want the kernel to believe it can count EL2 - because one day we may want to update KVM to allow the guest to count it's hypervisor overhead (e.g. host kernel time associated with the guest).

If we were to support this in the future, then exclude_hv will suddenly start meaning something in a guest, so this could be considered to be an ABI break.

Quoted 16 lines

I could write some preface that describes this outlook. Alternatively I could just spell out what happens on a guest, e.g.

"For the host this attribute will exclude EL1 and additionally EL2 on a VHE system.

For the guest this attribute will exclude EL1."

Though I'm less comfortable with this, as the last statement "For the guest this attribute will exclude EL1." describes the product of both kvm_pmu_set_counter_event_type and armv8pmu_set_event_filter which is confusing to work out and also makes an assumption that we don't have nested virt (true for now at least) and also reasons about bare-metal EL levels which probably aren't that useful for someone changing this logic or understanding what the flags do for there performance analysis.

Quoted 1 line

Do you have a preference for how this is improved?

I think you should be explicit about what is counted. If we don't count EL2 when profiling in a guest (regardless of the exclude_*) flags, then we should say that. By not documenting this we don't actually buy ourselves room to change things in future, we should have an emergent behaviour which isn't covered by our docs.

Will

#21 lore raw

Re: [PATCH v12 8/8] arm64: docs: document perf event attributes

Quoted 57 lines · Will Deacon
Quoted 31 lines · Andrew Murray
Quoted 14 lines · Will Deacon

Documentation/arm64/perf.txt line 23

Quoted 9 lines · Andrew Murray
Documentation/arm64/perf.txt new file +9 -0
+exclude_kernel
+--------------
+
+This attribute excludes the kernel.
+
+The kernel runs at EL2 with VHE and EL1 without. Guest kernels always run
+at EL1.
+
+This attribute will exclude EL1 and additionally EL2 on a VHE system.

I find this last sentence a bit confusing, because it can be read to imply that if you don't set exclude_kernel and you're in a guest on a VHE system, then you can profile EL2.

Yes this could be misleading.

However from the perspective of the guest, when exclude_kernel is not set we do indeed allow the guest to program it's PMU with ARMV8_PMU_INCLUDE_EL2 - and thus the statement above is correct in terms of what the kernel believes it is doing.

I think these statements are less confusing if we treat the exception levels as those 'detected' by the running context (e.g. consider the impact of nested virt here) - and we if ignore what the hypervisor (KVM) does outside (e.g. stops counting upon switching between guest/host, translating PMU filters in kvm_pmu_set_counter_event_type etc, etc). This then makes this document useful for those wishing to change this logic (which is the intent) rather than those trying to understand how we filter for EL levels as seen bare-metal.

Quoted 4 lines

With regards to the example you gave (exclude_kernel, EL2) - yes we want the kernel to believe it can count EL2 - because one day we may want to update KVM to allow the guest to count it's hypervisor overhead (e.g. host kernel time associated with the guest).

If we were to support this in the future, then exclude_hv will suddenly start meaning something in a guest, so this could be considered to be an ABI break.

Quoted 16 lines

I could write some preface that describes this outlook. Alternatively I could just spell out what happens on a guest, e.g.

"For the host this attribute will exclude EL1 and additionally EL2 on a VHE system.

For the guest this attribute will exclude EL1."

Though I'm less comfortable with this, as the last statement "For the guest this attribute will exclude EL1." describes the product of both kvm_pmu_set_counter_event_type and armv8pmu_set_event_filter which is confusing to work out and also makes an assumption that we don't have nested virt (true for now at least) and also reasons about bare-metal EL levels which probably aren't that useful for someone changing this logic or understanding what the flags do for there performance analysis.

Quoted 7 lines
Quoted 1 line

Do you have a preference for how this is improved?

I think you should be explicit about what is counted. If we don't count EL2 when profiling in a guest (regardless of the exclude_*) flags, then we should say that. By not documenting this we don't actually buy ourselves room to change things in future, we should have an emergent behaviour which isn't covered by our docs.

OK no problem, I'll update this.

Andrew Murray

Quoted 2 lines

Will