Re: [PATCH v1 01/14] xen/pci: Refactor MSI code that implements MSI functionality within XEN

[Julien Grall <julien@xen.org>]

On 19/08/2021 15:16, Rahul Singh wrote:
> Hi Julien,

Hi Rahul,

Sorry for the late reply.

>> On 19 Aug 2021, at 1:18 pm, Julien Grall <julien@xen.org> wrote:
>>
>> Hi Rahul,
>>
>> On 19/08/2021 13:02, Rahul Singh wrote:
>>> MSI code that implements MSI functionality to support MSI within XEN is
>>> not usable on ARM. Move the code under CONFIG_HAS_PCI_MSI flag to gate
>>
>> Can you clarify what you mean by not usable? Is it because we lack of support or we have no plan to use it?
> 
> We have no plan to use it. Code moved to CONFIG_HAS_PCI_MSI will only be required for ARM if we
I read "not usable", as the code doesn't yet work on Arm. This is quite 
possible, but from you wrote the main reason is that the code is not 
necessary yet (see below) on Arm.

> decide to support PCI device access (PCI MSI interrupt support) within XEN. As of now, we are planning
> to add support for PCI device access for DOM0/DOMU guests not for XEN.

This is probably going to happen sooner rather than later as, AFAICT, 
there are SMMUs out which signals using MSIs.

But, AFAICT, this code would also be used if we need to manage the MSIs 
in Xen on behalf of the guest (such as if the platform is using GICv2m).

That said, I would be fine with gating the MSI code behind a new config. 
However, I think the commit message wants to be clarified into why we 
don't need the option. Maybe something like:

"On Arm, the inital plan is to only support GICv3 ITS which doesn't 
require us to manage the MSIs because the HW will protect against 
spoofing. Introduce a new option XXXX to protect the MSI code.
"

Cheers,

-- 
Julien Grall