Re: [PATCH 3/3] riscv_cbqri: Bind to RQSC platform devices
From: Drew Fustini
Date: Thu Nov 19 2026 - 19:06:05 EST
On Wed, Nov 18, 2026 at 11:49:04PM +0000, Jonathan Cameron wrote:
> On Tue, 17 Nov 2026 14:11:29 -0800
> "Drew Fustini (Tenstorrent)" <fustini@xxxxxxxxxx> wrote:
>
> > Let the capacity controller driver probe the "cbqri-capacity" platform
> > devices created from the RQSC table. The cache is named by its PPTT
> > Cache ID and the PPTT gives its level. Select ACPI_RQSC so the table is
> > parsed and drop the OF dependency.
> >
> > No change for device tree systems.
> >
> > Signed-off-by: Drew Fustini (Tenstorrent) <fustini@xxxxxxxxxx>
>
> Has the DT support landed? If not feels like a few tweaks in that
> would make this less noisy.
>
> If not, see below for one tidy up I'd make here if you can't push
> it back to that series.
Sure, I will roll the changes around cbqri_capacity_of_cache() into a
v13 and leave that out of this patch.
> You almost never want to do manual device_node puts these days!
>
> It's one of the nicest places to use the cleanup.h magic.
Good idea, I will use it for device_node.
> > cache_np = of_parse_phandle(dev->of_node, "riscv,cbqri-cache", 0);
> Whilst you are here.
> struct device_node *cache_np __free(device_node) =
> of_parse_phandle();
> and drop all the gotos. I haven't checked but if there is
> a path where you want to hang on to it, no_free_ptr() that.
Will do. There should not be a need for no_free_ptr() for this one.
Thanks,
Drew