Re: [PATCH 3/5] riscv: dts: add initial T-HEAD light SoC device tree

["Heiko Stübner" <heiko@sntech.de>]

Am Montag, 8. Mai 2023, 18:44:04 CEST schrieb Conor Dooley:
> On Tue, May 09, 2023 at 12:26:10AM +0800, Jisheng Zhang wrote:
> > On Sun, May 07, 2023 at 10:35:12PM +0100, Conor Dooley wrote:
> > > On Mon, May 08, 2023 at 02:23:02AM +0800, Jisheng Zhang wrote:
> > > 
> > > > +		c910_0: cpu@0 {
> > > > +			compatible = "thead,c910", "riscv";
> > > > +			device_type = "cpu";
> > > > +			riscv,isa = "rv64imafdc";
> > > 
> > > Does this support more than "rv64imafdc"?
> > > I assume there's some _xtheadfoo extensions that it does support,
> > > although I am not sure how we are proceeding with those - Heiko might
> > > have a more nuanced take.
> > > 
> > > > +		reset: reset-sample {
> > > > +			compatible = "thead,reset-sample";
> > > 
> > > What is a "reset-sample"?
> > 
> > This node is only for opensbi. The compatible string is already in
> > opensbi. Do we also need to add dt-binding for it in linux?
> 
> If it's to be included in the kernel's dts, then yes, you do need a
> dt-binding. If you remove it, then you don't :)
> 
> That said, "thead,reset-sample" is a strangely named compatible, so if
> you do keep it it may end up needing a rename!

and you'll need to justify that this describes actual hardware
(dt-maintainers iterate all the time that dt is a hardware description, not
a configuration scheme).

The question also would be if this is part of upstream opensbi at all.


In general though, openSBI does something similar with their perf-counter
description. Describing the mapping and eventids usable in which counter
via a structure passed from u-boot to openSBI.

The difference here is, that openSBI then removes the relevant nodes from
the dt, so that the kernel never sees them [0] .

As you reset-sample seems to fall into a similar category, I guess
it would be better suited in an a foo-u-boot.dtsi ?


[0] https://github.com/riscv-software-src/opensbi/blob/master/lib/utils/fdt/fdt_pmu.c#L42