Re: [PATCH RFT 1/2] arm64: dts: toshiba: Update SoC and PCIe ranges to reflect hardware behavior

[Frank Li <Frank.li@nxp.com>]

On Mon, Jul 28, 2025 at 05:38:25AM +0000, nobuhiro1.iwamatsu@toshiba.co.jp wrote:
> Hi Frank,
>
> Thanks for your patch, and sorry reply was too late.
>
> > -----Original Message-----
> > From: Frank Li <Frank.Li@nxp.com>
> > Sent: Saturday, June 14, 2025 6:33 AM
> > To: iwamatsu nobuhiro(岩松 信洋 □ＤＩＴＣ○ＣＰＴ)
> > <nobuhiro1.iwamatsu@toshiba.co.jp>; Rob Herring <robh@kernel.org>;
> > Krzysztof Kozlowski <krzk+dt@kernel.org>; Conor Dooley
> > <conor+dt@kernel.org>; Lorenzo Pieralisi <lpieralisi@kernel.org>; Krzysztof
> > Wilczyński <kwilczynski@kernel.org>; Manivannan Sadhasivam
> > <mani@kernel.org>; Bjorn Helgaas <bhelgaas@google.com>
> > Cc: linux-arm-kernel@lists.infradead.org; devicetree@vger.kernel.org;
> > linux-kernel@vger.kernel.org; linux-pci@vger.kernel.org; Frank Li
> > <Frank.Li@nxp.com>
> > Subject: [PATCH RFT 1/2] arm64: dts: toshiba: Update SoC and PCIe ranges to
> > reflect hardware behavior
> >
> > tmpv7708 trim address bit[31:30] in tmpv7708 before passing to the PCIe
> > controller. So add a 'ranges' entry under the parent bus 'soc' to map address 0x0
> > to 0x40000000.
> >
> > Update the PCIe node's 'config' and 'ranges' properties to use the real upstream
> > bus address.
> >
> > Ensure there is no functional impact on the final address translation result.
> >
> > Prepare for the removal of the driver’s cpu_addr_fixup().
> >
> > Signed-off-by: Frank Li <Frank.Li@nxp.com>
> > ---
> >  arch/arm64/boot/dts/toshiba/tmpv7708.dtsi | 16 ++++++++++++----
> >  1 file changed, 12 insertions(+), 4 deletions(-)
> >
> > diff --git a/arch/arm64/boot/dts/toshiba/tmpv7708.dtsi
> > b/arch/arm64/boot/dts/toshiba/tmpv7708.dtsi
> > index 39806f0ae5133..2a18aa93d4723 100644
> > --- a/arch/arm64/boot/dts/toshiba/tmpv7708.dtsi
> > +++ b/arch/arm64/boot/dts/toshiba/tmpv7708.dtsi
> > @@ -147,7 +147,15 @@ soc {
> >  		#size-cells = <2>;
> >  		compatible = "simple-bus";
> >  		interrupt-parent = <&gic>;
> > -		ranges;
> > +		ranges = /* register 1:1 map */
> > +			 <0x0 0x24000000 0x0 0x24000000 0x0 0x10000000>,
> > +			 /*
> > +			  * bus fabric mask address bit 30 and 31 to 0
> > +			  * before send to PCIe controller.
> > +			  *
> > +			  * PCIe map address 0 to cpu's 0x40000000
> > +			  */
> > +			 <0x0 0x00000000 0x0 0x40000000 0x0 0x40000000>;
> >
> >  		gic: interrupt-controller@24001000 {
> >  			compatible = "arm,gic-400";
> > @@ -481,7 +489,7 @@ pwm: pwm@241c0000 {
> >  		pcie: pcie@28400000 {
> >  			compatible = "toshiba,visconti-pcie";
> >  			reg = <0x0 0x28400000 0x0 0x00400000>,
> > -			      <0x0 0x70000000 0x0 0x10000000>,
> > +			      <0x0 0x30000000 0x0 0x10000000>,
>
> If my understanding is correct, this setting conflicts with the address space this patch changed
> ranges above. Therefore, it does not work.
>
> 0x24000000 + 0x10000000 > 0x30000000

You are right.  Address map should only happen at pci fabic.

So we should not touch soc's ranges.

soc {
	...

	pci-bus {
		 ranges = /* register 1:1 map */
                 <0x0 0x24000000 0x0 0x24000000 0x0 0x10000000>,
                 /*
                  * bus fabric mask address bit 30 and 31 to 0
                  * before send to PCIe controller.
                  *
                  * PCIe map address 0 to cpu's 0x40000000
                  */
                 <0x0 0x00000000 0x0 0x40000000 0x0 0x40000000>;


		pci@0x24000000 {
			...
		}
	}
}

Do you need me send v2 for your test?

Frank
>
> By reducing 0x10000000 to 0xc000000, it will fit within the 0x30000000 range.
> And by adding the following to the driver:
> ```
> pci->use_parent_dt_ranges = true;
> ```
>
> the PCIe will work, but this setting prevents access to devices located after 0x30000000.
> Is there any other DT method to avoid this?
>
> Best regards,
>   Nobuhiro