[PATCH v2 2/8] arm/arm64: KVM: arch_timer: Only schedule soft timer on vcpu_block

[Eric Auger]

Hi Christoffer,
On 09/04/2015 09:40 PM, Christoffer Dall wrote:
> We currently schedule a soft timer every time we exit the guest if the
> timer did not expire while running the guest.  This is really not
> necessary, because the only work we do in the timer work function is to
> kick the vcpu.
> 
> Kicking the vcpu does two things:
> (1) If the vpcu thread is on a waitqueue, make it runnable and remove it
> from the waitqueue.
> (2) If the vcpu is running on a different physical CPU from the one
> doing the kick, it sends a reschedule IPI.
> 
> The second case cannot happen, because the soft timer is only ever
> scheduled when the vcpu is not running.  The first case is only relevant
> when the vcpu thread is on a waitqueue, which is only the case when the
> vcpu thread has called kvm_vcpu_block().
> 
> Therefore, we only need to make sure a timer is scheduled for
> kvm_vcpu_block(), which we do by encapsulating all calls to
> kvm_vcpu_block() with kvm_timer_{un}schedule calls.
> 
> Additionally, we only schedule a soft timer if the timer is enabled and
> unmasked, since it is useless otherwise.
> 
> Note that theoretically userspace can use the SET_ONE_REG interface to
> change registers that should cause the timer to fire, even if the vcpu
> is blocked without a scheduled timer, but this case was not supported
> before this patch and we leave it for future work for now.
> 
> Signed-off-by: Christoffer Dall <christoffer.dall at linaro.org>
> ---
>  arch/arm/include/asm/kvm_host.h   |  3 --
>  arch/arm/kvm/arm.c                | 10 +++++
>  arch/arm64/include/asm/kvm_host.h |  3 --
>  include/kvm/arm_arch_timer.h      |  2 +
>  virt/kvm/arm/arch_timer.c         | 91 ++++++++++++++++++++++++++-------------
>  5 files changed, 72 insertions(+), 37 deletions(-)
> 
> diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h
> index 86fcf6e..dcba0fa 100644
> --- a/arch/arm/include/asm/kvm_host.h
> +++ b/arch/arm/include/asm/kvm_host.h
> @@ -236,7 +236,4 @@ static inline void kvm_arm_setup_debug(struct kvm_vcpu *vcpu) {}
>  static inline void kvm_arm_clear_debug(struct kvm_vcpu *vcpu) {}
>  static inline void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu) {}
>  
> -static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
> -static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
> -
>  #endif /* __ARM_KVM_HOST_H__ */
> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
> index ce404a5..bdf8871 100644
> --- a/arch/arm/kvm/arm.c
> +++ b/arch/arm/kvm/arm.c
> @@ -271,6 +271,16 @@ int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
>  	return kvm_timer_should_fire(vcpu);
>  }
>  
> +void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
> +{
> +	kvm_timer_schedule(vcpu);
> +}
> +
> +void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
> +{
> +	kvm_timer_unschedule(vcpu);
> +}
> +
>  int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
>  {
>  	/* Force users to call KVM_ARM_VCPU_INIT */
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index dd143f5..415938d 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -257,7 +257,4 @@ void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
>  void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
>  void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
>  
> -static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
> -static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
> -
>  #endif /* __ARM64_KVM_HOST_H__ */
> diff --git a/include/kvm/arm_arch_timer.h b/include/kvm/arm_arch_timer.h
> index e1e4d7c..ef14cc1 100644
> --- a/include/kvm/arm_arch_timer.h
> +++ b/include/kvm/arm_arch_timer.h
> @@ -71,5 +71,7 @@ u64 kvm_arm_timer_get_reg(struct kvm_vcpu *, u64 regid);
>  int kvm_arm_timer_set_reg(struct kvm_vcpu *, u64 regid, u64 value);
>  
>  bool kvm_timer_should_fire(struct kvm_vcpu *vcpu);
> +void kvm_timer_schedule(struct kvm_vcpu *vcpu);
> +void kvm_timer_unschedule(struct kvm_vcpu *vcpu);
>  
>  #endif
> diff --git a/virt/kvm/arm/arch_timer.c b/virt/kvm/arm/arch_timer.c
> index 48c6e1a..7991537 100644
> --- a/virt/kvm/arm/arch_timer.c
> +++ b/virt/kvm/arm/arch_timer.c
> @@ -111,14 +111,21 @@ static enum hrtimer_restart kvm_timer_expire(struct hrtimer *hrt)
>  	return HRTIMER_NORESTART;
>  }
>  
> +static bool kvm_timer_irq_can_fire(struct kvm_vcpu *vcpu)
> +{
> +	struct arch_timer_cpu *timer = &vcpu->arch.timer_cpu;
> +
> +	return !(timer->cntv_ctl & ARCH_TIMER_CTRL_IT_MASK) &&
> +		(timer->cntv_ctl & ARCH_TIMER_CTRL_ENABLE) &&
> +		!kvm_vgic_get_phys_irq_active(timer->map);
kvm_vgic_get_phys_irq_active(timer->map) checks a logical state and not
the actual HW state. What is the exact aim of that check? in case the
PPI already is active, ie. timer hit, no use to schedule anything?

> +}
> +
>  bool kvm_timer_should_fire(struct kvm_vcpu *vcpu)
>  {
>  	struct arch_timer_cpu *timer = &vcpu->arch.timer_cpu;
>  	cycle_t cval, now;
>  
> -	if ((timer->cntv_ctl & ARCH_TIMER_CTRL_IT_MASK) ||
> -	    !(timer->cntv_ctl & ARCH_TIMER_CTRL_ENABLE) ||
> -	    kvm_vgic_get_phys_irq_active(timer->map))
> +	if (!kvm_timer_irq_can_fire(vcpu))
>  		return false;
>  
>  	cval = timer->cntv_cval;
> @@ -127,24 +134,61 @@ bool kvm_timer_should_fire(struct kvm_vcpu *vcpu)
>  	return cval <= now;
>  }
>  
> -/**
> - * kvm_timer_flush_hwstate - prepare to move the virt timer to the cpu
> - * @vcpu: The vcpu pointer
> - *
> - * Disarm any pending soft timers, since the world-switch code will write the
> - * virtual timer state back to the physical CPU.
> +/*
> + * Schedule the background timer before calling kvm_vcpu_block, so that this
> + * thread is removed from its waitqueue and made runnable when there's a timer
> + * interrupt to handle.
>   */
> -void kvm_timer_flush_hwstate(struct kvm_vcpu *vcpu)
> +void kvm_timer_schedule(struct kvm_vcpu *vcpu)
>  {
>  	struct arch_timer_cpu *timer = &vcpu->arch.timer_cpu;
> +	u64 ns;
> +	cycle_t cval, now;
> +
> +	BUG_ON(timer_is_armed(timer));
> +
> +	/*
> +	 * No need to schedule a background timer if the guest timer has
> +	 * already expired, because kvm_vcpu_block will return before putting
> +	 * the thread to sleep.
> +	 */
> +	if (kvm_timer_should_fire(vcpu))
> +		return;
>  
>  	/*
> -	 * We're about to run this vcpu again, so there is no need to
> -	 * keep the background timer running, as we're about to
> -	 * populate the CPU timer again.
> +	 * If the timer is either not capable of raising interrupts (disabled
> +	 * or masked) or if we already have a background timer, then there's
> +	 * no more work for us to do.
I don't understand the comment about "if we already have a background
timer", related to the above comment...
>  	 */
> +	if (!kvm_timer_irq_can_fire(vcpu))
> +		return;
> +
> +	/*  The timer has not yet expired, schedule a background timer */
> +	cval = timer->cntv_cval;
> +	now = kvm_phys_timer_read() - vcpu->kvm->arch.timer.cntvoff;
> +
> +	ns = cyclecounter_cyc2ns(timecounter->cc,
> +				 cval - now,
> +				 timecounter->mask,
> +				 &timecounter->frac);
> +	timer_arm(timer, ns);
> +}
> +
> +void kvm_timer_unschedule(struct kvm_vcpu *vcpu)
> +{
> +	struct arch_timer_cpu *timer = &vcpu->arch.timer_cpu;
>  	timer_disarm(timer);
> +}
>  
> +/**
> + * kvm_timer_flush_hwstate - prepare to move the virt timer to the cpu
> + * @vcpu: The vcpu pointer
> + *
> + * Check if the virtual timer has expired while we were running in the host,
> + * and inject an interrupt if that was the case.
> + */
> +void kvm_timer_flush_hwstate(struct kvm_vcpu *vcpu)
> +{
>  	/*
>  	 * If the timer expired while we were not scheduled, now is the time
>  	 * to inject it.
above comment seems duplicated now?
> @@ -157,32 +201,17 @@ void kvm_timer_flush_hwstate(struct kvm_vcpu *vcpu)
>   * kvm_timer_sync_hwstate - sync timer state from cpu
>   * @vcpu: The vcpu pointer
>   *
> - * Check if the virtual timer was armed and either schedule a corresponding
> - * soft timer or inject directly if already expired.
> + * Check if the virtual timer has expired while we were running in the guest,
> + * and inject an interrupt if that was the case.
>   */
>  void kvm_timer_sync_hwstate(struct kvm_vcpu *vcpu)
>  {
>  	struct arch_timer_cpu *timer = &vcpu->arch.timer_cpu;
> -	cycle_t cval, now;
> -	u64 ns;
>  
>  	BUG_ON(timer_is_armed(timer));
>  
> -	if (kvm_timer_should_fire(vcpu)) {
> -		/*
> -		 * Timer has already expired while we were not
> -		 * looking. Inject the interrupt and carry on.
> -		 */
> +	if (kvm_timer_should_fire(vcpu))
>  		kvm_timer_inject_irq(vcpu);
> -		return;
> -	}
> -
> -	cval = timer->cntv_cval;
> -	now = kvm_phys_timer_read() - vcpu->kvm->arch.timer.cntvoff;
> -
> -	ns = cyclecounter_cyc2ns(timecounter->cc, cval - now, timecounter->mask,
> -				 &timecounter->frac);
> -	timer_arm(timer, ns);
>  }
>  
>  int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu,
>