linux-kernel - Re: [PATCH v2 1/4] KVM: arm/arm64: vgic: Do not cond_resched

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] [day] [month] [year] [list]

Message-ID: <0573c8b2-db76-19e3-db76-5433b2e4ad0a@arm.com>
Date:   Fri, 14 Dec 2018 09:36:59 +0000
From:   Julien Thierry <julien.thierry@....com>
To:     Christoffer Dall <christoffer.dall@....com>
Cc:     linux-kernel@...r.kernel.org, kvmarm@...ts.cs.columbia.edu,
        marc.zyngier@....com, linux-arm-kernel@...ts.infradead.org,
        linux-rt-users@...r.kernel.org, tglx@...utronix.de,
        rostedt@...dmis.org, bigeasy@...utronix.de, stable@...r.kernel.org
Subject: Re: [PATCH v2 1/4] KVM: arm/arm64: vgic: Do not cond_resched_lock()
 with IRQs disabled



On 11/12/2018 10:20, Christoffer Dall wrote:
> On Mon, Nov 26, 2018 at 06:26:44PM +0000, Julien Thierry wrote:
>> To change the active state of an MMIO, halt is requested for all vcpus of
>> the affected guest before modifying the IRQ state. This is done by calling
>> cond_resched_lock() in vgic_mmio_change_active(). However interrupts are
>> disabled at this point and we cannot reschedule a vcpu.
>>
>> Solve this by waiting for all vcpus to be halted after emmiting the halt
>> request.
>>
>> Signed-off-by: Julien Thierry <julien.thierry@....com>
>> Suggested-by: Marc Zyngier <marc.zyngier@....com>
>> Cc: Christoffer Dall <christoffer.dall@....com>
>> Cc: Marc Zyngier <marc.zyngier@....com>
>> Cc: stable@...r.kernel.org
>> ---
>>  virt/kvm/arm/vgic/vgic-mmio.c | 36 ++++++++++++++----------------------
>>  1 file changed, 14 insertions(+), 22 deletions(-)
>>
>> diff --git a/virt/kvm/arm/vgic/vgic-mmio.c b/virt/kvm/arm/vgic/vgic-mmio.c
>> index f56ff1c..5c76a92 100644
>> --- a/virt/kvm/arm/vgic/vgic-mmio.c
>> +++ b/virt/kvm/arm/vgic/vgic-mmio.c
>> @@ -313,27 +313,6 @@ static void vgic_mmio_change_active(struct kvm_vcpu *vcpu, struct vgic_irq *irq,
>>  
>>  	spin_lock_irqsave(&irq->irq_lock, flags);
>>  
>> -	/*
>> -	 * If this virtual IRQ was written into a list register, we
>> -	 * have to make sure the CPU that runs the VCPU thread has
>> -	 * synced back the LR state to the struct vgic_irq.
>> -	 *
>> -	 * As long as the conditions below are true, we know the VCPU thread
>> -	 * may be on its way back from the guest (we kicked the VCPU thread in
>> -	 * vgic_change_active_prepare)  and still has to sync back this IRQ,
>> -	 * so we release and re-acquire the spin_lock to let the other thread
>> -	 * sync back the IRQ.
>> -	 *
>> -	 * When accessing VGIC state from user space, requester_vcpu is
>> -	 * NULL, which is fine, because we guarantee that no VCPUs are running
>> -	 * when accessing VGIC state from user space so irq->vcpu->cpu is
>> -	 * always -1.
>> -	 */
>> -	while (irq->vcpu && /* IRQ may have state in an LR somewhere */
>> -	       irq->vcpu != requester_vcpu && /* Current thread is not the VCPU thread */
>> -	       irq->vcpu->cpu != -1) /* VCPU thread is running */
>> -		cond_resched_lock(&irq->irq_lock);
>> -
>>  	if (irq->hw) {
>>  		vgic_hw_irq_change_active(vcpu, irq, active, !requester_vcpu);
>>  	} else {
>> @@ -368,8 +347,21 @@ static void vgic_mmio_change_active(struct kvm_vcpu *vcpu, struct vgic_irq *irq,
>>   */
>>  static void vgic_change_active_prepare(struct kvm_vcpu *vcpu, u32 intid)
>>  {
>> -	if (intid > VGIC_NR_PRIVATE_IRQS)
>> +	if (intid > VGIC_NR_PRIVATE_IRQS) {
>> +		struct kvm_vcpu *tmp;
>> +		int i;
>> +
>>  		kvm_arm_halt_guest(vcpu->kvm);
>> +
>> +		/* Wait for each vcpu to be halted */
>> +		kvm_for_each_vcpu(i, tmp, vcpu->kvm) {
>> +			if (tmp == vcpu)
>> +				continue;
>> +
>> +			while (tmp->cpu != -1)
>> +				cond_resched();
>> +		}
> 
> I'm actually thinking we don't need this loop at all after the requet
> rework which causes:
> 
>  1. kvm_arm_halt_guest() to use kvm_make_all_cpus_request(kvm, KVM_REQ_SLEEP), and
>  2. KVM_REQ_SLEEP uses REQ_WAIT, and
>  3. REQ_WAIT requires the VCPU to respond to IPIs before returning, and
>  4. a VCPU thread can only respond when it enables interrupt, and
>  5. enabling interrupts when running a VCPU only happens after syncing
>     the VGIC hwstate.
> 
> Does that make sense?

I'm not super familiar with what goes on with the vgic hwstate syncing,
but looking at kvm_arm_halt_guest() and kvm_arch_vcpu_ioctl_run(), I
agree with the reasoning.

> It would be good if someone can validate this, but if it holds this
> patch just becomes a nice deletion of the logic in
> vgic-mmio_change_active.
> 

As long as running kvm_vgic_sync_hwstate() on each vcpu is all that is
needed before we can modify the active state, I think your solution is
definitely the way to go.

Thanks,

-- 
Julien Thierry