Re: [PATCH-4.5 4/4] xen/arm: set GICH_HCR_NPIE if all the LRs are in use

[Julien Grall <julien.grall@linaro.org>]

On 02/10/2014 04:59 PM, Stefano Stabellini wrote:
> On Fri, 7 Feb 2014, Julien Grall wrote:
>> On 07/02/14 18:56, Stefano Stabellini wrote:
>>> On return to guest, if there are no free LRs and we still have more
>>> interrupt to inject, set GICH_HCR_NPIE so that we are going to receive a
>>> maintenance interrupt when no pending interrupts are present in the LR
>>> registers.
>>> The maintenance interrupt handler won't do anything anymore, but
>>> receiving the interrupt is going to cause gic_inject to be called on
>>> return to guest that is going to clear the old LRs and inject new
>>> interrupts.
>>>
>>> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
>>> ---
>>>   xen/arch/arm/gic.c |    8 +++++++-
>>>   1 file changed, 7 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/xen/arch/arm/gic.c b/xen/arch/arm/gic.c
>>> index 87bd5d3..bee2618 100644
>>> --- a/xen/arch/arm/gic.c
>>> +++ b/xen/arch/arm/gic.c
>>> @@ -810,8 +810,14 @@ void gic_inject(void)
>>>       gic_restore_pending_irqs(current);
>>>       if (!gic_events_need_delivery())
>>>           gic_inject_irq_stop();
>>> -    else
>>> +    else {
>>>           gic_inject_irq_start();
>>> +    }
>>> +
>>> +    if ( !list_empty(&current->arch.vgic.lr_pending) )
>>> +        GICH[GICH_HCR] = GICH_HCR_EN | GICH_HCR_NPIE;
>>> +    else
>>> +        GICH[GICH_HCR] = GICH_HCR_EN;
>>
>> Any reason to not move this in the else?
> 
> Yes: we need to be able to disable GICH_HCR_NPIE even if there are no
> irqs to inject.

In this case, can we simply enable/disable GICH_HCR_NPIE instead of
resetting the register every time? GICH_HCR contains other bits that
could be set in the future.

>> BTW, I think we can safely avoid to reinject the IRQ for the event channels if
>> there is pending events. It will also avoid spurious IRQ on some specific case
>> :).
> 
> Currently we don't set GIC_IRQ_GUEST_PENDING for evtchn_irq if the irq
> is already inflight. I'll see if I can come up with a patch to
> streamline that behaviour.

Ok.


-- 
Julien Grall