Hi Anthony,
>>
>> Yes I can change it, a need to do it anyway because I think also here, the
>> suggestion
>> From Jan can apply and we could pass a negative value that means “max VL
>> supported
>> by the platform"
>
> Well, it's a config file, not a C ABI, so max allowed here doesn't have to be
> spelled "-1", it could also be "max", "max-allowed",
> "max-size-supported", ... So fill free deviate from the restricted C
> ABI. But "-1" works as long as it's the only allowed negative number.
Yes while working on the patch I’ve found that I could declare this type in
Libxl:
libxl_sve_type = Enumeration("sve_type", [
(0, "disabled"),
(128, "128"),
(256, "256"),
(384, "384"),
(512, "512"),
(640, "640"),
(768, "768"),
(896, "896"),
(1024, "1024"),
(1152, "1152"),
(1280, "1280"),
(1408, "1408"),
(1536, "1536"),
(1664, "1664"),
(1792, "1792"),
(1920, "1920"),
(2048, "2048"),
(-1, "hw")
], init_val = "LIBXL_SVE_TYPE_DISABLED”)
So that in xl I can just use libxl_sve_type_from_string
>
>>>
>>>> +supported bits value, then the domain creation will fail.
>>>> +A value equal to zero is the default and it means this guest is not
>>>> allowed to
>>>> +use SVE.
>>>> +
>>>> +=back
>>>> +
>>>> =head3 x86
>>>>
>>>> =over 4
>>>> diff --git a/tools/libs/light/libxl_arm.c b/tools/libs/light/libxl_arm.c
>>>> index ddc7b2a15975..16a49031fd51 100644
>>>> --- a/tools/libs/light/libxl_arm.c
>>>> +++ b/tools/libs/light/libxl_arm.c
>>>> @@ -211,6 +211,8 @@ int libxl__arch_domain_prepare_config(libxl__gc *gc,
>>>> return ERROR_FAIL;
>>>> }
>>>>
>>>> + config->arch.sve_vl = d_config->b_info.arch_arm.sve;
>>>
>>> This truncate a 16bit value into an 8bit value, I think you should check
>>> that the value can actually fit.
>>>
>>> And maybe check `d_config->b_info.arch_arm.sve` value here instead of
>>> `xl` as commented later.
>>
>> Yes I can do it, one question, can I use here xc_physinfo to retrieve the
>> maximum
>> Vector length from arch_capabilities?
>> I mean, is there a better way or I can go for that?
>
> Yeah, there might be a "better" way. I think me suggestion to check the
> sve value here was wrong. I still want to have it checked in libxl, but
> it might be better to do that in the previous step, that is
> "libxl__domain_config_setdefault". libxl__arch_domain_build_info_setdefault()
> will have `physinfo` so you won't have to call xc_physinfo().
Right, I’ve seen it before but I was unsure if it was the right way, now that
you
suggested it, I will go for that.
Thank you.
Cheers,
Luca
