Re: [PATCH v1 12/27] xen/riscv: introduce aia_init() and aia_available()

[Oleksii Kurochko <oleksii.kurochko@gmail.com>]

On 4/16/26 2:06 PM, Jan Beulich wrote:
> On 13.04.2026 11:32, Oleksii Kurochko wrote:
>> On 4/2/26 11:00 AM, Jan Beulich wrote:
>>> On 10.03.2026 18:08, Oleksii Kurochko wrote:
>>>> aia_init() is going to contain all the stuff related to AIA initialization.
>>>> At the moment, it is just Check if SSAIA extension is available and if yes
>>>> set is_aia_available to true.
>>>
>>> And (future) users of aia_available() can't directly call
>>> riscv_isa_extension_available()? Nor can aia_available() be a convenience
>>> wrapper around that call? It's only ...
>>>
>>>> --- /dev/null
>>>> +++ b/xen/arch/riscv/aia.c
>>>> @@ -0,0 +1,25 @@
>>>> +/* SPDX-License-Identifier: GPL-2.0-only */
>>>> +
>>>> +#include <xen/errno.h>
>>>> +#include <xen/init.h>
>>>> +#include <xen/sections.h>
>>>> +#include <xen/types.h>
>>>> +
>>>> +#include <asm/cpufeature.h>
>>>> +
>>>> +static bool __ro_after_init is_aia_available;
>>>
>>> ... a boolean, yes, but still.
>>
>> My purpose was to have a variable which represent that AIA is
>> initialized properly. Maybe, it makes sense to rename this variable to
>> is_aia_inited.
> 
> I'm generally having trouble with the non-word "inited". How about
> "aia_usable"? Fits ...
> 
>> The idea is that in future patches VGEIN will be also initialized in
>> aia_init() and if wasn't initialized properly then just keep
>> is_aia_availabe be set to false and in such case we will have that
>> is_aia_available != riscv_isa_extension_available(NULL,
>> RISCV_ISA_EXT_ssaia).
> 
> ... this as well. And ...
> 
>> Note that regarding VGEIN it is arguable that it should be initialized
>> as generally it could be that there is no VGEIN what means that h/w
>> assisted guest interrupt files aren't available and s/w one should be
>> used. But s/w guest interrupt files aren't supported.
>> So I mean that with the current implementation if VGEIN isn't
>> initialized I will tell that AIA isn't available what generally isn't
>> quite true.
> 
> ... this. Question of course if what aia_usable (or whatever its name)
> being set to false means to Xen's own operation. If Xen would still work
> okay (without being able to create guests), perhaps the name may want to
> be "aia_guest_usable" or some such. I.e. as I said on a number of
> occasions: A variable wants to be named to accurately express its
> purpose, without the name growing excessively long.

aia_usable sounds good to me.

It covers also the case if AIA isn't expected to be used as h/w uses 
non-AIA-compliant interrupt controller, for example, PLIC.

~ Oleksii