Re: [PATCH 2/2] riscv: dts: spacemit: k3: Add PCIe device node

[Anirudh Srinivasan]

Hi Inochi,

On Sun, Jul 12, 2026 at 11:19 PM Inochi Amaoto <inochiama@xxxxxxxxx> wrote:
>
> On Sun, Jul 12, 2026 at 11:13:07PM -0500, Anirudh Srinivasan wrote:
> > Hi Inochi,
> >
> > On Thu, Jul 09, 2026 at 12:04:14PM +0800, Inochi Amaoto wrote:
> > > Add all PCIe device node for Spacemit K3.
> > >
> > > Signed-off-by: Inochi Amaoto <inochiama@xxxxxxxxx>
> > > ---
> > >  arch/riscv/boot/dts/spacemit/k3-pico-itx.dts |  38 ++++
> > >  arch/riscv/boot/dts/spacemit/k3-pinctrl.dtsi |  33 ++++
> > >  arch/riscv/boot/dts/spacemit/k3.dtsi         | 195 +++++++++++++++++++
> > >  3 files changed, 266 insertions(+)
> > >
> >
> > I see some warnings printed during probe
> >
> > [   10.842955] r8169 0002:01:00.0: of_irq_parse_pci: failed with rc=-22
> >
> > [   10.937754] rtw89_8852be 0004:01:00.0: of_irq_parse_pci: failed with rc=-22
> >
>
> This is an expected error because the driver does not support INTx
> but the hardware claims it supports. Since the vendor does not want
> to support it, so there is no way for the upstreamer to add this.

Are you saying that the hardware advertises support for INTx, but
spacemit hasn't provided a driver that uses it?

Is there a way to let the driver know to not try setting up legacy
interrupts then? It seems like it's trying to do that now, and failing
to find interrupts node in the DT.

>
> > I see that all the 3 rc nodes are missing an interrupts node and an
> > interrupt parent (saplic). The vendor DT seems to have this. Not sure if
> > this is related.
> >
> > The 10G ethernet probes fine, but the wifi card fails to probe. Seems
> > like the card is trying to request 32 bit DMA addresses and we're
> > failing cause we have 64 bit memory only.
> > The vendor kernel seems to have a patch to enable 36 bit DMA in
> > the rtw89 driver. That'll probably fix this.
> >
>
> This warning is also found on my test, it is not related to
> the PCIe driver, but a lack of the RTL phy driver. :(
>
> You may also noticed a bad performance on this 10G ethernet,
> it is caused by the single queue used by the RTL driver.
> IIRC, even 9800 X3D can not achieve the max speed.

I don't have the ability to test the 10G NIC unfortunately :(

Regards
Anirudh Srinivasan

>
> Regards,
> Inochi
>