Re: [PATCH v3 08/12] xen/riscv: rework G-stage mode handling

[Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>]

On 4/21/26 11:39 AM, Jan Beulich wrote:
> On 10.04.2026 17:54, Oleksii Kurochko wrote:
> > --- a/xen/arch/riscv/p2m.c
> > +++ b/xen/arch/riscv/p2m.c
> > @@ -45,12 +45,27 @@ struct p2m_pte_ctx {
> >       unsigned int level;          /* Paging level at which the PTE resides. */
> >   };
> >   
> > -static struct gstage_mode_desc __ro_after_init max_gstage_mode = {
> > -    .mode = HGATP_MODE_OFF,
> > -    .paging_levels = 0,
> > -    .name = "Bare",
> > +/* Values should be sorted by ->mode in this array */
> > +static const struct gstage_mode_desc modes[] = {
> As before, I'm of the clear opinion that this is too generic an identifier
> for use at file scope.
I can rename it to gstage_modes or p2m_modes to make it better suite 
file scope.
> > --- a/xen/include/public/arch-riscv.h
> > +++ b/xen/include/public/arch-riscv.h
> > @@ -56,6 +56,11 @@ typedef struct vcpu_guest_context vcpu_guest_context_t;
> >   DEFINE_XEN_GUEST_HANDLE(vcpu_guest_context_t);
> >   
> >   struct xen_arch_domainconfig {
> > +    /*
> > +     * G-stage MMU mode for the guest (e.g. "sv39", "sv48", "sv57").
> > +     * Must be set; an empty string is invalid.
> > +     */
> > +    char gstage_mode[8];
> >   };
> I have to say that I find it odd to use a string literal for this purpose.
> Specifying the number of wanted address bits would feel more natural. Plus
> the strings named aren't valid G-stage modes afaict - they lack the x4.
I will change it to unsigned char gstage_mode and put there on of the 
value of HGATP_MODE_SV39X4, HGATP_MODE_SV48X4. Or just address bits as 
you suggested will be fine too.
~ Oleksii