Subject: Re: [PATCH 3/3] KVM: x86: do not scan IRR twice on APICv vmentry
To: Paolo Bonzini <pbonzini@redhat.com>, linux-kernel@vger.kernel.org,
        kvm@vger.kernel.org
References: <1475011213-34225-1-git-send-email-pbonzini@redhat.com>
 <1475011213-34225-4-git-send-email-pbonzini@redhat.com>
Cc: feng.wu@intel.com, mst@redhat.com, rkrcmar@redhat.com
From: Yang Zhang <yang.zhang.wz@gmail.com>
Message-ID: <30768946-c4f6-f9bd-5495-2a3adefffc69@gmail.com>
Date: Fri, 14 Oct 2016 15:32:34 +0800
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:45.0) Gecko/20100101
 Thunderbird/45.4.0
MIME-Version: 1.0
In-Reply-To: <1475011213-34225-4-git-send-email-pbonzini@redhat.com>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 7497
Lines: 213

On 2016/9/28 5:20, Paolo Bonzini wrote:
> Calling apic_find_highest_irr results in IRR being scanned twice,
> once in vmx_sync_pir_from_irr and once in apic_search_irr.  Change
> vcpu_enter_guest to use sync_pir_from_irr (with a new argument to
> trigger the RVI write), and let sync_pir_from_irr get the new RVI
> directly from kvm_apic_update_irr.
>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  arch/x86/include/asm/kvm_host.h |  2 +-
>  arch/x86/kvm/lapic.c            | 25 ++++++++++++++++---------
>  arch/x86/kvm/lapic.h            |  4 ++--
>  arch/x86/kvm/svm.c              |  2 +-
>  arch/x86/kvm/vmx.c              | 25 ++++++++++++++-----------
>  arch/x86/kvm/x86.c              |  7 +++----
>  6 files changed, 37 insertions(+), 28 deletions(-)
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 4b20f7304b9c..f73eea243567 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -941,7 +941,7 @@ struct kvm_x86_ops {
>  	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
>  	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
>  	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
> -	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
> +	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu, bool sync_rvi);
>  	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
>  	int (*get_tdp_level)(void);
>  	u64 (*get_mt_mask)(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio);
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index be8b7ad56dd1..85b79b90e3d0 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -317,7 +317,7 @@ static int find_highest_vector(void *bitmap)
>  	     vec >= 0; vec -= APIC_VECTORS_PER_REG) {
>  		reg = bitmap + REG_POS(vec);
>  		if (*reg)
> -			return fls(*reg) - 1 + vec;
> +			return __fls(*reg) + vec;
>  	}
>
>  	return -1;
> @@ -337,25 +337,32 @@ static u8 count_vectors(void *bitmap)
>  	return count;
>  }
>
> -void __kvm_apic_update_irr(u32 *pir, void *regs)
> +int __kvm_apic_update_irr(u32 *pir, void *regs)
>  {
> -	u32 i, pir_val;
> +	u32 i, vec;
> +	u32 pir_val, irr_val;
> +	int max_irr = -1;
>
> -	for (i = 0; i <= 7; i++) {
> +	for (i = vec = 0; i <= 7; i++, vec += 32) {

how about ignore the first 32 vectors since they cannot be used as 
normal interrupt except nmi interrupt which is special handled.

>  		pir_val = READ_ONCE(pir[i]);
> +		irr_val = *((u32 *)(regs + APIC_IRR + i * 0x10));
>  		if (pir_val) {
> -			pir_val = xchg(&pir[i], 0);
> -			*((u32 *)(regs + APIC_IRR + i * 0x10)) |= pir_val;
> +			irr_val |= xchg(&pir[i], 0);
> +			*((u32 *)(regs + APIC_IRR + i * 0x10)) = irr_val;
>  		}
> +		if (irr_val)
> +			max_irr = __fls(irr_val) + vec;
>  	}
> +
> +	return max_irr;
>  }
>  EXPORT_SYMBOL_GPL(__kvm_apic_update_irr);
>
> -void kvm_apic_update_irr(struct kvm_vcpu *vcpu, u32 *pir)
> +int kvm_apic_update_irr(struct kvm_vcpu *vcpu, u32 *pir)
>  {
>  	struct kvm_lapic *apic = vcpu->arch.apic;
>
> -	__kvm_apic_update_irr(pir, apic->regs);
> +	return __kvm_apic_update_irr(pir, apic->regs);
>  }
>  EXPORT_SYMBOL_GPL(kvm_apic_update_irr);
>
> @@ -376,7 +383,7 @@ static inline int apic_find_highest_irr(struct kvm_lapic *apic)
>  		return -1;
>
>  	if (apic->vcpu->arch.apicv_active)
> -		kvm_x86_ops->sync_pir_to_irr(apic->vcpu);
> +		kvm_x86_ops->sync_pir_to_irr(apic->vcpu, false);
>  	result = apic_search_irr(apic);

You have got the max_irr in sync_pir_to_irr. It seems the 
apic_search_irr() is redundant here.

>  	ASSERT(result == -1 || result >= 16);

I think the result must >= 32. But it seems this code exists when kvm 
first merging into kernel.

>
> diff --git a/arch/x86/kvm/lapic.h b/arch/x86/kvm/lapic.h
> index f60d01c29d51..fc72828c3782 100644
> --- a/arch/x86/kvm/lapic.h
> +++ b/arch/x86/kvm/lapic.h
> @@ -70,8 +70,8 @@ int kvm_lapic_reg_read(struct kvm_lapic *apic, u32 offset, int len,
>  bool kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
>  			   int short_hand, unsigned int dest, int dest_mode);
>
> -void __kvm_apic_update_irr(u32 *pir, void *regs);
> -void kvm_apic_update_irr(struct kvm_vcpu *vcpu, u32 *pir);
> +int __kvm_apic_update_irr(u32 *pir, void *regs);
> +int kvm_apic_update_irr(struct kvm_vcpu *vcpu, u32 *pir);
>  int kvm_apic_set_irq(struct kvm_vcpu *vcpu, struct kvm_lapic_irq *irq,
>  		     struct dest_map *dest_map);
>  int kvm_apic_local_deliver(struct kvm_lapic *apic, int lvt_type);
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 9b4221471e3d..60e53fa2b554 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -4383,7 +4383,7 @@ static void svm_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap)
>  	return;
>  }
>
> -static void svm_sync_pir_to_irr(struct kvm_vcpu *vcpu)
> +static void svm_sync_pir_to_irr(struct kvm_vcpu *vcpu, bool sync_rvi)
>  {
>  	return;
>  }
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 207b9aa32915..1edefab54d01 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -4852,17 +4852,6 @@ static void vmx_deliver_posted_interrupt(struct kvm_vcpu *vcpu, int vector)
>  		kvm_vcpu_kick(vcpu);
>  }
>
> -static void vmx_sync_pir_to_irr(struct kvm_vcpu *vcpu)
> -{
> -	struct vcpu_vmx *vmx = to_vmx(vcpu);
> -
> -	if (!pi_test_on(&vmx->pi_desc) ||
> -	    !pi_test_and_clear_on(&vmx->pi_desc))
> -		return;
> -
> -	kvm_apic_update_irr(vcpu, vmx->pi_desc.pir);
> -}
> -
>  /*
>   * Set up the vmcs's constant host-state fields, i.e., host-state fields that
>   * will not change in the lifetime of the guest.
> @@ -8609,6 +8598,20 @@ static void vmx_hwapic_irr_update(struct kvm_vcpu *vcpu, int max_irr)
>  	}
>  }
>
> +static void vmx_sync_pir_to_irr(struct kvm_vcpu *vcpu, bool sync_rvi)
> +{
> +	struct vcpu_vmx *vmx = to_vmx(vcpu);
> +	int max_irr;
> +
> +	if (!pi_test_on(&vmx->pi_desc) ||
> +	    !pi_test_and_clear_on(&vmx->pi_desc))
> +		return;
> +
> +	max_irr = kvm_apic_update_irr(vcpu, vmx->pi_desc.pir);
> +	if (sync_rvi)
> +		vmx_hwapic_irr_update(vcpu, max_irr);
> +}
> +
>  static void vmx_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap)
>  {
>  	if (!kvm_vcpu_apicv_active(vcpu))
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 604cfbfc5bee..148e14fdc55d 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -2821,7 +2821,7 @@ static int kvm_vcpu_ioctl_get_lapic(struct kvm_vcpu *vcpu,
>  				    struct kvm_lapic_state *s)
>  {
>  	if (vcpu->arch.apicv_active)
> -		kvm_x86_ops->sync_pir_to_irr(vcpu);
> +		kvm_x86_ops->sync_pir_to_irr(vcpu, false);
>
>  	return kvm_apic_get_state(vcpu, s);
>  }
> @@ -6448,7 +6448,7 @@ static void vcpu_scan_ioapic(struct kvm_vcpu *vcpu)
>  		kvm_scan_ioapic_routes(vcpu, vcpu->arch.ioapic_handled_vectors);
>  	else {
>  		if (vcpu->arch.apicv_active)
> -			kvm_x86_ops->sync_pir_to_irr(vcpu);
> +			kvm_x86_ops->sync_pir_to_irr(vcpu, false);
>  		kvm_ioapic_scan_entry(vcpu, vcpu->arch.ioapic_handled_vectors);
>  	}
>  	bitmap_or((ulong *)eoi_exit_bitmap, vcpu->arch.ioapic_handled_vectors,
> @@ -6611,8 +6611,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
>  		 * virtual interrupt delivery.
>  		 */
>  		if (vcpu->arch.apicv_active)
> -			kvm_x86_ops->hwapic_irr_update(vcpu,
> -				kvm_lapic_find_highest_irr(vcpu));
> +			kvm_x86_ops->sync_pir_to_irr(vcpu, true);
>  	}
>
>  	if (kvm_check_request(KVM_REQ_EVENT, vcpu) || req_int_win) {
>


-- 
Yang
Alibaba Cloud Computing