Re: [PART2 PATCH v3 10/11] svm: Introduce AMD IOMMU avic_ga_log_notifier

[Suravee Suthikulpanit]

Hi Radim,

Again, sorry for possible confusion from the mis-use of the word ga_tag 
for parameter naming in different places. Please see my reply for the 
patch 07/11 for more detail.

On 07/12/2016 02:09 AM, Radim KrÄmÃÅ wrote:
> 2016-07-11 05:11-0500, Suravee Suthikulpanit:
> > From: Suravee Suthikulpanit <Suravee.Suthikulpanit@xxxxxxx>
> >
> > This patch introduces avic_ga_log_notifier, which will be called
> > by IOMMU driver whenever it handles the Guest vAPIC (GA) log entry.
> >
> > Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@xxxxxxx>
> > ---
> > diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> > @@ -928,6 +930,45 @@ static void svm_disable_lbrv(struct vcpu_svm *svm)
> > +static int avic_ga_log_notifier(int avic_tag, int vcpu_id, int vec)
> > +{
> > +	unsigned long flags;
> > +	struct kvm_arch *ka = NULL;
> > +	struct kvm_vcpu *vcpu = NULL;
> > +	struct vcpu_svm *svm = NULL;
> > +
> > +	pr_debug("SVM: %s: avic_tag=%#x, vcpu_id=%#x, vec=%#x\n",
> > +		 __func__, avic_tag, vcpu_id, vec);
> > +
> > +	spin_lock_irqsave(&svm_vm_data_hash_lock, flags);
> > +	hash_for_each_possible(svm_vm_data_hash, ka, hnode, avic_tag) {
> > +		struct kvm *kvm = container_of(ka, struct kvm, arch);
>
> iommu_handle_ga_guest_nr_entry(), the caller, already performed a hash
> lookup to get ir_data.  Cannot we store the vcpu pointer in ir_data, to
> avoid a second hash lookup and a vcpu lookup?
>
> Thanks.

The iommu_handle_ga_guest_nr_entry() has the look up logic to map ga_tag 
back to an ir_data in order to see if there are any pending interrupts 
targeting a particular VCPU of a particular VM.

The look up in the avic_ga_log_notifier() is for SVM to map VM_ID to the 
struct kvm_arch.

Thanks,
Suravee