Re: [PATCH v1 05/27] xen/riscv: introduce guest riscv,isa string

[Jan Beulich <jbeulich@suse.com>]

On 10.04.2026 12:24, Oleksii Kurochko wrote:
> On 4/1/26 3:49 PM, Jan Beulich wrote:
>> On 10.03.2026 18:08, Oleksii Kurochko wrote:
>>> +#define MAX_GUEST_ISA_STR_LEN 256
>>> +char guest_isa_str[MAX_GUEST_ISA_STR_LEN];
>>
>> __ro_after_init?
>>
>> Yet then - can this really be a global? Isn't the set of extensions
>> available to a guest a per-guest property, i.e. a global could at best
>> represent an upper bound on features?
> 
> At least, for now I think it is okay just to use global guest riscv,isa 
> for all the domains. From my point of view if Xen doesn't support some 
> extension to be used by guest so then should be responsible to generate 
> proper riscv,isa.
> 
> It probably make sense to have riscv,isa per guest but then we still 
> will want to parse this passed riscv,isa and check what Xen supports and 
> what not, what looks like a double work a little bit. Probably it is the 
> best one option and if riscv,isa isn't provided then just use that one 
> generated by Xen.

My take is: Have data arrangement be as it wants to be in the end (i.e.
here: record what ought to be per guest as per-guest data). All guests
using the same value for now is a fair restriction until things have
progressed more.

Jan