Re: [PATCH v1 05/14] xen/riscv: introduce platform_get_irq()

[Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>]

On 4/15/25 1:23 PM, Jan Beulich wrote:

> On 15.04.2025 13:11, Oleksii Kurochko wrote:
> > On 4/10/25 5:35 PM, Jan Beulich wrote:
> > > On 08.04.2025 17:57, Oleksii Kurochko wrote:
> > > > @@ -21,6 +22,22 @@ static struct intc_info __ro_after_init aplic_info = {
> > > >       .hw_version = INTC_APLIC,
> > > >   };
> > > >   
> > > > +static int aplic_irq_xlate(const uint32_t *intspec, unsigned int intsize,
> > > >           
> > > As you start adding functions calling indirectly, please consider adding cf_check
> > > right away, even if right now this has no effect on RISC-V. That'll save you from
> > > going through the entire RISC-V subtree later on to find them all.
> > >         
> > Sure. I thought that it is a feature for x86 as I haven't seen such attribute for
> > Arm and RISC-V in GCC manuals.
> >       
> And that looks to be correct. I was under the (admittedly vague) impression
> Arm64 had something equivalent in hardware, which then merely needs enabling
> in the compiler. Not sure about RISC-V, but seeing the endless flow of
> patches enabling new extensions in binutils, it would perhaps even be
> surprising if nothing along these lines was already in the works somewhere.
>     
You are right, something is already in the work:
- https://github.com/riscv/riscv-cfi
- https://lore.kernel.org/lkml/20230213045351.3945824-1-debug@xxxxxxxxxxxx/ (interesting that
  they are enabling it for U-mode)


~ Oleksii