Re: [RFC 09/14] xen/xsm: flask: MSI is PCI specific

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

Hi Ian,

On 03/17/2014 10:13 AM, Ian Campbell wrote:
> On Fri, 2014-03-14 at 18:15 +0000, Julien Grall wrote:
>> Hi Ian,
>>
>> On 03/14/2014 05:25 PM, Ian Campbell wrote:
>>> On Thu, 2014-03-13 at 14:40 +0000, Julien Grall wrote:
>>>>
>>>> Right, nr_static_irqs is set to 1024 on ARM. For now every IRQs, are not
>>>> higher than this number.
>>>
>>> This will change before long though. Would HAVE_MSI_IRQS be a better
>>> name for this particular arch-configurable? After all PCI without MSI is
>>> plausible isn't it? (well, maybe not in the modern world)
>>
>> Most of PCI passthrough code is relying that MSI is also implemented.
>> Futhermore, I think the plan for ARM is to support both PCI and MSI.
> 
> Right, but we might get PCI support first and separately I think?

Good question, I though it's better to come with PCI and MSI support at
the time.

>> That's why I chose HAS_PCI.
> 
> I suppose we can revisit it easily enough if it causes problems.

If we implement first PCI, we might need to modify a couple of other
place in Xen. I think we can delay HAS_MSI_IRQS and see if we really
need it.

Regards,

-- 
Julien Grall