Re: [PATCH v3] arm: irq: increase size of irq from uint8_t to uint32_t

[Julien Grall <julien.grall@citrix.com>]

Hi Ian,

On 15/04/15 13:10, Ian Campbell wrote:
> Switching Julien to his Citrix address which should probably be used in
> the future.
> 
> On Wed, 2015-04-08 at 17:14 +0300, Iurii Konovalenko wrote:
>> From: Iurii Konovalenko <iurii.konovalenko@globallogic.com>
>>
>> Changes are dedicated to XEN_DOMCTL_irq_permission and
>> IRQ pssthrough API functions.
>>
>> PHYSDEV_* operations already using 32 bits type but signed one.
>>
>> Although, PHYSDEV_* operations are not yet used on ARM and LPIs support
>> (which are using very high number) are not supported yet, we don't need
>> to care about theses for now.
> 
> I may have slightly lost track, but I think we decided in Julien's
> passthrough thread not to use most of these interfaces on ARM, or am I
> confused?

Only XEN_DOMCTL_bind_pt_irq will be used for ARM passthrough.

I asked Iurii to mention PHYSDEV_* because the prototype is already
valid but use int rather than unsigned int.

> If I'm correct then I think we can avoid messing with many of other
> ones, for example the ISA IRQ one doesn't need changing, does it? (ISA
> only had 16 IRQs IIRC...)

Only the newly introduce function xc_domain_bind_pt_spi_irq will be used
for ARM. I guess we can avoid to modify xc_domain_bind_pt*.

> I think it would be best if whichever bits of this are still relevant
> were folded into Julien's '[PATCH v5 p2 04/19] xen/arm: Implement
> hypercall DOMCTL_{,un}bind_pt_pirq' or at least presented as a followup
> to it.
> 
>> diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
>> index 8803ab2..65fb866 100644
>> --- a/xen/include/public/domctl.h
>> +++ b/xen/include/public/domctl.h
>> @@ -400,7 +400,7 @@ DEFINE_XEN_GUEST_HANDLE(xen_domctl_setdebugging_t);
>>  
>>  /* XEN_DOMCTL_irq_permission */
>>  struct xen_domctl_irq_permission {
>> -    uint8_t pirq;
>> +    uint32_t pirq;
>>      uint8_t allow_access;    /* flag to specify enable/disable of IRQ access */
> 
> I think we weren't going to end up using this one either, but again I
> might not be remembering correctly.

Even though we won't support it on ARM for now, I think it's good to
keep the interface consistent.

It would avoid us to forget the problem when this will be support it later.

Regards,

-- 
Julien Grall