Re: [PATCH v1 13/27] xen/riscv: add basic VGEIN management for AIA guests

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

On 4/17/26 4:07 PM, Jan Beulich wrote:
> On 17.04.2026 13:34, Oleksii Kurochko wrote:
>> On 4/16/26 2:21 PM, Jan Beulich wrote:
>>> On 13.04.2026 16:42, Oleksii Kurochko wrote:
>>>> On 4/2/26 12:03 PM, Jan Beulich wrote:
>>>>> On 10.03.2026 18:08, Oleksii Kurochko wrote:
>>>>>> @@ -14,12 +27,109 @@ bool aia_available(void)
>>>>>>         return is_aia_available;
>>>>>>     }
>>>>>>     
>>>>>> +int __init vgein_init(unsigned int cpu)
>>>>>
>>>>> If this needs invoking once per CPU being brought up, it can't be __init.
>>>>
>>>> Yes, it is going to be called inside the secondary CPU bring-up function.
>>>>
>>>> __init sections are removed much later, after all CPUs are brought up,
>>>> so it looks like that at the moment when secondary CPUs are being
>>>> brought up, __init still exists and can be called.
>>>
>>> No. And I said so before. See XEN_SYSCTL_cpu_hotplug (and ignore it having
>>> "hotplug" in the name, but merely consider that you can take CPUs offline
>>> at runtime, and later bring them online again).
>>
>> Thanks, now it makes sense to drop __init.
>>
>>>>> Also - static?
>>>>
>>>> It isn't static because it will be called inside the secondary CPU
>>>> bring-up function.
>>>
>>> As it doesn't need calling from the outside for the boot CPU, it's not
>>> obvious why it would need calling from the outside for secondary ones.
>>
>> I think I am confused here by what do you mean by "calling from the
>> outside".
> 
> As in "from another translation unit".
> 
>> It should be called during boot or secondary CPU initialization (if AIA
>> is used), right?
>> Then considering that VGEIN-related stuff is connected to AIA then
>> vgein_init() is expected to live in aia.c and considering that it wants
>> to be called by secondary CPU boot code it can't be static.
> 
> Putting my question differently: Why is there a difference here between
> boot CPU and secondary CPUs. My (general) expectation would be that either
> the function wants calling from outside of aia.c in all cases, or it could
> be static (now and later).

Generally, if update prototype of aia_init() to receive cpuid argument 
then it will be enough to make vgein_init() static and then all others 
will call aia_init(). The only reason why I tried to have vgein_init() 
non-static is that in the case of secondary CPU there is no need to call 
second time anything (at least, now) from aia_init() or re-init 
is_aia_usable variable.

But I can agree that probably it would be better to make vgein_init() 
static now and then use aia_init() everywhere it is needed just for the 
case that if something new will be added what should be called for boot 
and secondary CPU won't be missed to call properly. And I think it is 
fine just re-init is_aia_usable without any additional synchronization.
As an option I could check at the start if is_aia_usable is initialized 
by checking if it is true and if it is then just call vgein_init() for 
that cpuid.
Would it be better?

> 
>>>>>> +{
>>>>>> +    struct vgein_bmp *vgein = &per_cpu(vgein_bmp, cpu);
>>>>>> +
>>>>>> +    csr_write(CSR_HGEIE, -1UL);
>>>>>> +    vgein->geilen = flsl(csr_read(CSR_HGEIE));
>>>>>> +    csr_write(CSR_HGEIE, 0);
>>>>>> +    if ( vgein->geilen )
>>>>>> +        vgein->geilen--;
>>>>>
>>>>> I don't understand this. The "len" in "geilen" stands for "length", I suppose,
>>>>> i.e. the number of bits. Hmm, the spec itself is inconsistent: "The number of
>>>>> bits implemented in hgeip and hgeie for guest external interrupts is UNSPECIFIED
>>>>> and may be zero. This number is known as GEILEN." This may or may not include
>>>>> bit 0 (which is implemented, but r/o zero). Then saying "Hence, if GEILEN is
>>>>> nonzero, bits GEILEN:1 shall be writable in ..." suggests 0 isn't included, but
>>>>> that's not unambiguous.
>>>>
>>>> But they explicitly wrote that: The least-significant bits are
>>>> implemented first, apart from bit 0. So bit 0 is explicitly excluded.
>>>
>>> Fine, but not including it in anything named *LEN feels - as said - ambiguous.
>>
>> Agree, it is. I just tried to follow the spec naming here. But I am okay
>> to rename it to max_gein, for example.
> 
> Perhaps better stick to spec naming, but consider adding a clarifying comment.

I will add then above geilen field in vgein_bmp struct the comment:
  The least-significant bits are implemented first, apart from bit 0

> 
>>>>>> +    /*
>>>>>> +     * All vCPU guest interrupt files are used and we don't support a case
>>>>>> +     * when number of vCPU on 1 pCPU is bigger then geilen.
>>>>>> +     */
>>>>>
>>>>> This wants checking in vgein_init() then. CPUs (beyond the boot one)
>>>>> violating this should not be brought online.
>>>>
>>>> It'll be nice. But we can't know how many vCPUs will be ran on pCPU when
>>>> vgein_init() is executed.
>>>
>>> I don't understand: How does it matter how many vCPU-s will exist later on
>>> (and will run wherever)?
>>
>> IIUC, you want to check what is mentioned in the comment in vcpu_init().
>> The comment says that it is checking that number of vCPU on one pCPU
>> isn't bigger then geilen. To check that we have to know an amount of
>> vCPU potentially will be ran on pCPU.
> 
> Well, not quite. I don't understand at all why the number of vCPU-s in
> a system would be limited by any pCPU property. I don't think we have
> any such constraint on x86, for example.

Then I will just drop this comment and correspondent ASSERT() and just 
return vgein_id in the case when there is no available h/w VGEINs to 
mark that h/w should use s/w guest interrupt files:

     if ( vgein_id > vgein->geilen )
         vgein_id = 0;
     else
         __set_bit(vgein_id, bmp);

~ Oleksii