kvm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Shannon Zhao <shannon.zhao@linaro.org>
To: Marc Zyngier <marc.zyngier@arm.com>,
	Shannon Zhao <zhaoshenglong@huawei.com>,
	kvmarm@lists.cs.columbia.edu, christoffer.dall@linaro.org
Cc: kvm@vger.kernel.org, will.deacon@arm.com,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH v12 16/21] KVM: ARM64: Add PMU overflow interrupt routing
Date: Tue, 23 Feb 2016 22:52:16 +0800	[thread overview]
Message-ID: <56CC7220.3020807@linaro.org> (raw)
In-Reply-To: <56CC695E.4010805@arm.com>



On 2016/2/23 22:14, Marc Zyngier wrote:
> On 22/02/16 09:37, Shannon Zhao wrote:
>> From: Shannon Zhao <shannon.zhao@linaro.org>
>>
>> When calling perf_event_create_kernel_counter to create perf_event,
>> assign a overflow handler. Then when the perf event overflows, set the
>> corresponding bit of guest PMOVSSET register. If this counter is enabled
>> and its interrupt is enabled as well, kick the vcpu to sync the
>> interrupt.
>>
>> On VM entry, if there is counter overflowed, inject the interrupt with
>> the level set to 1. Otherwise, inject the interrupt with the level set
>> to 0.
>>
>> Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
>> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
>> Reviewed-by: Andrew Jones <drjones@redhat.com>
>
> As I mentioned yesterday, I was trying to pinpoint a performance drop, so I added PMU support to kvmtool (and made it an optional flag). This allowed be to find this:
>
>> ---
>>   arch/arm/kvm/arm.c    |  2 ++
>>   include/kvm/arm_pmu.h |  2 ++
>>   virt/kvm/arm/pmu.c    | 47 ++++++++++++++++++++++++++++++++++++++++++++++-
>>   3 files changed, 50 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
>> index dda1959..f54264c 100644
>> --- a/arch/arm/kvm/arm.c
>> +++ b/arch/arm/kvm/arm.c
>> @@ -28,6 +28,7 @@
>>   #include <linux/sched.h>
>>   #include <linux/kvm.h>
>>   #include <trace/events/kvm.h>
>> +#include <kvm/arm_pmu.h>
>>
>>   #define CREATE_TRACE_POINTS
>>   #include "trace.h"
>> @@ -577,6 +578,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
>>   		 * non-preemptible context.
>>   		 */
>>   		preempt_disable();
>> +		kvm_pmu_flush_hwstate(vcpu);
>>   		kvm_timer_flush_hwstate(vcpu);
>>   		kvm_vgic_flush_hwstate(vcpu);
>>
>> diff --git a/include/kvm/arm_pmu.h b/include/kvm/arm_pmu.h
>> index 8bc92d1..cf68f9a 100644
>> --- a/include/kvm/arm_pmu.h
>> +++ b/include/kvm/arm_pmu.h
>> @@ -44,6 +44,7 @@ u64 kvm_pmu_valid_counter_mask(struct kvm_vcpu *vcpu);
>>   void kvm_pmu_disable_counter(struct kvm_vcpu *vcpu, u64 val);
>>   void kvm_pmu_enable_counter(struct kvm_vcpu *vcpu, u64 val);
>>   void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val);
>> +void kvm_pmu_flush_hwstate(struct kvm_vcpu *vcpu);
>>   void kvm_pmu_software_increment(struct kvm_vcpu *vcpu, u64 val);
>>   void kvm_pmu_handle_pmcr(struct kvm_vcpu *vcpu, u64 val);
>>   void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data,
>> @@ -67,6 +68,7 @@ static inline u64 kvm_pmu_valid_counter_mask(struct kvm_vcpu *vcpu)
>>   static inline void kvm_pmu_disable_counter(struct kvm_vcpu *vcpu, u64 val) {}
>>   static inline void kvm_pmu_enable_counter(struct kvm_vcpu *vcpu, u64 val) {}
>>   static inline void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val) {}
>> +static inline void kvm_pmu_flush_hwstate(struct kvm_vcpu *vcpu) {}
>>   static inline void kvm_pmu_software_increment(struct kvm_vcpu *vcpu, u64 val) {}
>>   static inline void kvm_pmu_handle_pmcr(struct kvm_vcpu *vcpu, u64 val) {}
>>   static inline void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu,
>> diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
>> index cda869c..6ac52ce 100644
>> --- a/virt/kvm/arm/pmu.c
>> +++ b/virt/kvm/arm/pmu.c
>> @@ -21,6 +21,7 @@
>>   #include <linux/perf_event.h>
>>   #include <asm/kvm_emulate.h>
>>   #include <kvm/arm_pmu.h>
>> +#include <kvm/arm_vgic.h>
>>
>>   /**
>>    * kvm_pmu_get_counter_value - get PMU counter value
>> @@ -181,6 +182,49 @@ void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val)
>>   }
>>
>>   /**
>> + * kvm_pmu_flush_hwstate - flush pmu state to cpu
>> + * @vcpu: The vcpu pointer
>> + *
>> + * Inject virtual PMU IRQ if IRQ is pending for this cpu.
>> + */
>> +void kvm_pmu_flush_hwstate(struct kvm_vcpu *vcpu)
>> +{
>> +	struct kvm_pmu *pmu = &vcpu->arch.pmu;
>> +	u64 overflow;
>> +
>> +	if (!kvm_arm_pmu_v3_ready(vcpu))
>> +		return;
>> +
>> +	overflow = kvm_pmu_overflow_status(vcpu);
>> +	kvm_vgic_inject_irq(vcpu->kvm, vcpu->vcpu_id, pmu->irq_num, !!overflow);
>
> It turns out that this single line costs us about 400 cycles on each entry:
>
Oh, it's really a time-consuming function.

> maz@flakes:~/kvm-ws-tests$ make LKVM=~/kvmtool/lkvm LKVM_ARGS=--pmu PERF=perf_4.3 tests-gicv2
> GICv2:
> do_hvc.bin:5690.17
> do_sgi.bin:9395.05
> do_sysreg.bin:5912.6
> maz@flakes:~/kvm-ws-tests$ make LKVM=~/kvmtool/lkvm PERF=perf_4.3 tests-gicv2
> GICv2:
> do_hvc.bin:5285.02
> do_sgi.bin:9131.24
> do_sysreg.bin:5563.7
>
> Caching the irq state and only injecting it if it has changed (just like we do for the timer) brings performance back to its previous level:
>
Marc, thanks a lot for finding out this. I'll merge below changes into 
this series and send out v13 tomorrow.

> diff --git a/include/kvm/arm_pmu.h b/include/kvm/arm_pmu.h
> index 176913f..b23e636 100644
> --- a/include/kvm/arm_pmu.h
> +++ b/include/kvm/arm_pmu.h
> @@ -35,6 +35,7 @@ struct kvm_pmu {
>   	int irq_num;
>   	struct kvm_pmc pmc[ARMV8_PMU_MAX_COUNTERS];
>   	bool ready;
> +	bool irq_level;
>   };
>
>   #define kvm_arm_pmu_v3_ready(v)		((v)->arch.pmu.ready)
> diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
> index 904617e..7156f8b 100644
> --- a/virt/kvm/arm/pmu.c
> +++ b/virt/kvm/arm/pmu.c
> @@ -229,13 +229,17 @@ void kvm_pmu_overflow_set(struct kvm_vcpu *vcpu, u64 val)
>   void kvm_pmu_flush_hwstate(struct kvm_vcpu *vcpu)
>   {
>   	struct kvm_pmu *pmu = &vcpu->arch.pmu;
> -	u64 overflow;
> +	bool overflow;
>
>   	if (!kvm_arm_pmu_v3_ready(vcpu))
>   		return;
>
> -	overflow = kvm_pmu_overflow_status(vcpu);
> -	kvm_vgic_inject_irq(vcpu->kvm, vcpu->vcpu_id, pmu->irq_num, !!overflow);
> +	overflow = !!kvm_pmu_overflow_status(vcpu);
> +	if (pmu->irq_level != overflow) {
> +		pmu->irq_level = overflow;
> +		kvm_vgic_inject_irq(vcpu->kvm, vcpu->vcpu_id,
> +				    pmu->irq_num, overflow);
> +	}
>   }
>
>   static inline struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc)
>
>
> Thanks,
>
> 	M.
>

-- 
Shannon

  reply	other threads:[~2016-02-23 14:52 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-02-22  9:37 [PATCH v12 00/21] KVM: ARM64: Add guest PMU support Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 01/21] ARM64: Move PMU register related defines to asm/perf_event.h Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 02/21] KVM: ARM64: Define PMU data structure for each vcpu Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 03/21] KVM: ARM64: Add offset defines for PMU registers Shannon Zhao
2016-02-22 17:51   ` Marc Zyngier
2016-02-23  1:46     ` Shannon Zhao
2016-02-23  8:33       ` Marc Zyngier
2016-02-23  9:29       ` Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 04/21] KVM: ARM64: Add access handler for PMCR register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 05/21] KVM: ARM64: Add access handler for PMSELR register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 06/21] KVM: ARM64: Add access handler for PMCEID0 and PMCEID1 register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 07/21] KVM: ARM64: PMU: Add perf event map and introduce perf event creating function Shannon Zhao
2016-02-23 17:42   ` Marc Zyngier
2016-02-24  1:27     ` Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 08/21] KVM: ARM64: Add access handler for event type register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 09/21] KVM: ARM64: Add access handler for event counter register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 10/21] KVM: ARM64: Add access handler for PMCNTENSET and PMCNTENCLR register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 11/21] KVM: ARM64: Add access handler for PMINTENSET and PMINTENCLR register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 12/21] KVM: ARM64: Add access handler for PMOVSSET and PMOVSCLR register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 13/21] KVM: ARM64: Add access handler for PMSWINC register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 14/21] KVM: ARM64: Add helper to handle PMCR register bits Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 15/21] KVM: ARM64: Add access handler for PMUSERENR register Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 16/21] KVM: ARM64: Add PMU overflow interrupt routing Shannon Zhao
2016-02-23 14:14   ` Marc Zyngier
2016-02-23 14:52     ` Shannon Zhao [this message]
2016-02-22  9:37 ` [PATCH v12 17/21] KVM: ARM64: Reset PMU state when resetting vcpu Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 18/21] KVM: ARM64: Free perf event of PMU when destroying vcpu Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 19/21] KVM: ARM64: Add a new feature bit for PMUv3 Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 20/21] KVM: ARM: Introduce per-vcpu kvm device controls Shannon Zhao
2016-02-22  9:37 ` [PATCH v12 21/21] KVM: ARM64: Add a new vcpu device control group for PMUv3 Shannon Zhao

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=56CC7220.3020807@linaro.org \
    --to=shannon.zhao@linaro.org \
    --cc=christoffer.dall@linaro.org \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=marc.zyngier@arm.com \
    --cc=will.deacon@arm.com \
    --cc=zhaoshenglong@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).