linux-kernel - Re: [PATCH 2/4] KVM: x86: refactor PIT state inject

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <56B22E95.30504@redhat.com>
Date:	Wed, 3 Feb 2016 17:45:09 +0100
From:	Paolo Bonzini <pbonzini@...hat.com>
To:	Radim Krčmář <rkrcmar@...hat.com>,
	linux-kernel@...r.kernel.org
Cc:	kvm@...r.kernel.org, Yuki Shibuya <shibuya.yk@...s.nec.co.jp>
Subject: Re: [PATCH 2/4] KVM: x86: refactor PIT state inject_lock



On 03/02/2016 17:23, Radim Krčmář wrote:
> Following patches would be even uglier if inject_lock didn't go away.
> 
> Patch changes the virtual wire comment to better describe our situation.
> 
> Signed-off-by: Radim Krčmář <rkrcmar@...hat.com>
> ---
>  arch/x86/kvm/i8254.c | 67 ++++++++++++++++++++++------------------------------
>  arch/x86/kvm/i8254.h |  3 +--
>  2 files changed, 29 insertions(+), 41 deletions(-)
> 
> diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
> index de5f5018026f..a137eb381012 100644
> --- a/arch/x86/kvm/i8254.c
> +++ b/arch/x86/kvm/i8254.c
> @@ -236,16 +236,13 @@ static void kvm_pit_ack_irq(struct kvm_irq_ack_notifier *kian)
>  {
>  	struct kvm_kpit_state *ps = container_of(kian, struct kvm_kpit_state,
>  						 irq_ack_notifier);
> -	int value;
>  
> -	spin_lock(&ps->inject_lock);
> +	atomic_set(&ps->irq_ack, 1);

smp_mb__before_atomic();

>  	if (atomic_add_unless(&ps->pending, -1, 0))
>  		/* in this case, we had multiple outstanding pit interrupts
>  		 * that we needed to inject.  Reinject
>  		 */
>  		queue_kthread_work(&ps->pit->worker, &ps->pit->expired);
> -	ps->irq_ack = 1;
> -	spin_unlock(&ps->inject_lock);
>  }
>  
>  void __kvm_migrate_pit_timer(struct kvm_vcpu *vcpu)
> @@ -276,34 +273,25 @@ static void pit_do_work(struct kthread_work *work)
>  	struct kvm_vcpu *vcpu;
>  	int i;
>  	struct kvm_kpit_state *ps = &pit->pit_state;
> -	int inject = 0;
>  
> -	/* Try to inject pending interrupts when
> -	 * last one has been acked.
> +	if (!atomic_xchg(&ps->irq_ack, 0))
> +		return;
> +
> +	kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 1, false);
> +	kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 0, false);
> +
> +	/*
> +	 * Provides NMI watchdog support via Virtual Wire mode.
> +	 * The route is: PIT -> LVT0 in NMI mode.
> +	 *
> +	 * Note: Our Virtual Wire implementation does not follow
> +	 * the MP specification.  We propagate a PIT interrupt to all
> +	 * VCPUs and only when LVT0 is in NMI mode.  The interrupt can
> +	 * also be simultaneously delivered through PIC and IOAPIC.
>  	 */
> -	spin_lock(&ps->inject_lock);
> -	if (ps->irq_ack) {
> -		ps->irq_ack = 0;
> -		inject = 1;
> -	}
> -	spin_unlock(&ps->inject_lock);
> -	if (inject) {
> -		kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 1, false);
> -		kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 0, false);
> -
> -		/*
> -		 * Provides NMI watchdog support via Virtual Wire mode.
> -		 * The route is: PIT -> PIC -> LVT0 in NMI mode.
> -		 *
> -		 * Note: Our Virtual Wire implementation is simplified, only
> -		 * propagating PIT interrupts to all VCPUs when they have set
> -		 * LVT0 to NMI delivery. Other PIC interrupts are just sent to
> -		 * VCPU0, and only if its LVT0 is in EXTINT mode.
> -		 */
> -		if (atomic_read(&kvm->arch.vapics_in_nmi_mode) > 0)
> -			kvm_for_each_vcpu(i, vcpu, kvm)
> -				kvm_apic_nmi_wd_deliver(vcpu);
> -	}
> +	if (atomic_read(&kvm->arch.vapics_in_nmi_mode) > 0)
> +		kvm_for_each_vcpu(i, vcpu, kvm)
> +			kvm_apic_nmi_wd_deliver(vcpu);
>  }
>  
>  static enum hrtimer_restart pit_timer_fn(struct hrtimer *data)
> @@ -323,6 +311,12 @@ static enum hrtimer_restart pit_timer_fn(struct hrtimer *data)
>  		return HRTIMER_NORESTART;
>  }
>  
> +static void kvm_pit_reset_reinject(struct kvm_pit *pit)
> +{
> +	atomic_set(&pit->pit_state.pending, 0);

smp_wmb()?

Looks safe otherwise.  (Please also add a comment before the memory
barriers to show the pairing).

Paolo