On 2014/11/17 18:14, Marc Zyngier wrote:
> On 17/11/14 09:27, wanghaibin wrote:
>> When queue irq(whatever SGI, PPI, SPI) failed, that means the ICH_LRn is not
>> enough, so, if this circs happened, skip the follow irq queue logic, and flag
>> the HCR.UIE directly.
>>
>> Signed-off-by: wanghaibin <wanghaibin.wang@xxxxxxxxxx>
>> ---
>> virt/kvm/arm/vgic.c | 12 +++++++++---
>> 1 files changed, 9 insertions(+), 3 deletions(-)
>>
>> diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
>> index 8390475..12026e6 100644
>> --- a/virt/kvm/arm/vgic.c
>> +++ b/virt/kvm/arm/vgic.c
>> @@ -1415,20 +1415,26 @@ static void __kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
>>
>> /* SGIs */
>> for_each_set_bit(i, vgic_cpu->pending_percpu, VGIC_NR_SGIS) {
>> - if (!vgic_queue_sgi(vcpu, i))
>> + if (!vgic_queue_sgi(vcpu, i)) {
>> overflow = 1;
>> + goto epilog;
>> + }
>> }
>>
>> /* PPIs */
>> for_each_set_bit_from(i, vgic_cpu->pending_percpu, VGIC_NR_PRIVATE_IRQS) {
>> - if (!vgic_queue_hwirq(vcpu, i))
>> + if (!vgic_queue_hwirq(vcpu, i)) {
>> overflow = 1;
>> + goto epilog;
>> + }
>> }
>>
>> /* SPIs */
>> for_each_set_bit(i, vgic_cpu->pending_shared, vgic_nr_shared_irqs(dist)) {
>> - if (!vgic_queue_hwirq(vcpu, i + VGIC_NR_PRIVATE_IRQS))
>> + if (!vgic_queue_hwirq(vcpu, i + VGIC_NR_PRIVATE_IRQS)) {
>> overflow = 1;
>> + goto epilog;
>> + }
>> }
>>
>> epilog:
>>
>
> If you had taken some time to actually *read* the code, you would have
> noticed that even in an overflow condition, it is possible to queue more
> interrupts (hint: You can go from ACTIVE to ACTIVE+PENDING without
> allocating a new LR).
>
> You call this an optimization, I call it an optimization killer.
I am mindless, this is a wrong patch.
Sorry.
>
> M.
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
