Re: [PATCH v7 3/4] xen/riscv: panic if the device tree advertises neither Svade nor Svadu

[Baptiste Le Duc <baptiste.le-duc@xxxxxxxxxx>]

On 26.10.2026 10:32, Jan Beulich wrote:
> On 16.10.2026 17:34, Baptiste Le Duc wrote:
> > Suggested-by: Jan Beulich <jbeulich@xxxxxxxx>
> Did I, really? Iirc I left open whether to go the panic() route or the
> alternative one. Based on Andrew's clarification as to pre-Svad{e,u}
> behavior, I think I further tried to indicate that synthesizing the
> correct extension may be preferable, to allow older systems' DTs to be
> used as-is. In any event ...
No, you didn't, sorry. I misread your "The outlier is when DT lacks any
information" as "panic when DT lacks any information", while you had just
agreed to probing in exactly that case. I'll drop the tag.

For v8 I plan to:
- trust the DT whenever it advertises Svade and/or Svadu, without probing.
- probe only when it advertises neither, set the probed extension in
  riscv_isa, and print a warning asking for the DT to be updated.

Does that match what you had in mind?

> > +    if ( !riscv_isa_extension_available(NULL, RISCV_ISA_EXT_svade) &&
> > +         !riscv_isa_extension_available(NULL, RISCV_ISA_EXT_svadu) )
> > +        panic("Device tree advertises neither Svade nor Svadu, A/D bits update 
> > scheme unknown\n");
> ... this message leaves the observer to figure out what to do about the
> situation.
Agreed. With the above the panic() goes away, and the warning will name
the extension to add to "riscv,isa".

Thanks,
Baptiste