Re: [PATCH v1 1/2] xen/riscv: skip disabled CPU nodes when probing ISA extensions

[Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>]

On 11/9/26 2:23 PM, Baptiste Le Duc wrote:
> > riscv_fill_hwcap_from_isa_string() intersects the riscv,isa of every CPU
> > node under /cpus, including the ones with status = "disabled". Such a
> > hart is never used by Xen, yet it can still remove extensions from the
> > host ISA: e.g. a disabled hart without the H extension makes Xen panic
> > with "Xen requires extension: h", even though every hart Xen runs on
> > implements it.
> >
> > Skip CPU nodes which aren't available.
> has_isa_extensions_property() still walks the disabled nodes, so the two
> loops over /cpus now disagree on which harts count. Rather than adding
> the check loop by loop, can we filter them in one place, e.g. with a
> dt_for_each_available_child_node() iterator, so no /cpus walk sees a disabled
> hart?
I missed updating has_isa_extensions_property() because I had Cristian 
Andrei's patch in mind, which adds support for the riscv,isa-extensions 
property (the patch isn't published to ML). This function will be 
completely dropped as part of that work.
So, until that patch is published, it would probably be better to add an 
additional check to has_isa_extensions_property().
Regarding the question of whether to use 
dt_for_each_available_child_node(), I don't see many use cases for it 
beyond this particular one. There are also plenty of cases where it's 
perfectly fine for a node to be disabled and still parsed, so I'm not 
entirely convinced that we need this change now.
Let's just add an explicit check to has_isa_extensions_property() for 
now. Are you okay with that or still think it'll be better to introduce 
dt_for_each_available_child_node() specifically for this case?
~ Oleksii