Re: [PATCH v9 05/10] xen/arm: physical irq follow virtual irq

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

Hi Stefano,

On 01/08/14 18:22, Stefano Stabellini wrote:
> On Mon, 28 Jul 2014, Julien Grall wrote:
>> On 07/28/2014 05:20 PM, Stefano Stabellini wrote:
>>> On Mon, 28 Jul 2014, Julien Grall wrote:
>>>> Hi stefano,
>>>>
>>>> On 07/24/2014 06:33 PM, Stefano Stabellini wrote:
>>>>> +void arch_move_irqs(struct vcpu *v)
>>>>> +{
>>>>> +    const cpumask_t *cpu_mask = cpumask_of(v->processor);
>>>>> +    struct domain *d = v->domain;
>>>>> +    struct pending_irq *p;
>>>>> +    struct vcpu *v_target;
>>>>> +    int i;
>>>>> +
>>>>> +    for ( i = 32; i < d->arch.vgic.nr_lines; i++ )
>>>>
>>>> Sorry, I didn't spot this error until now.
>>>>
>>>> For the VGIC nr_lines contains the number of *SPIs* rather on the GIC
>>>> structure it's the number of IRQs... the name is very confusing. I have
>>>> a patch to rename nr_lines into nr_spis, along with adding a macro
>>>> vgic_number_lines.
>>>
>>> I couldn't parse this sentence.
>>
>> Sorry it was not very clear.
>>
>>> I guess you are saying that vgic.nr_lines doesn't represent the number
>>> of spis?
>>
>> Yes. In the VGIC structure nr_lines = number of SPIs.
>>
>> On GIC structure nr_lines = number of IRQs.
>
> Wait a second. If in the vgic struct, nr_lines = number of SPIs, then
> the code in this patch is correct as it is, isn't it?

No because your loop starts at 32 and you check against vgic.nr_lines.

The loop should be:

for ( i = 32; i < (d->arch.vgic.nr_lines + 32); i++ )

Otherwise you will forget to migrate the last 32 SPIs.

Regards,

-- 
Julien Grall