[PATCH v2 0/2] Add R8A77980/Condor PCIe support

[PATCH v2 0/2] Add R8A77980/Condor PCIe support

[Sergei Shtylyov]

Hello!

Here's the set of 2 patches against Simon Horman's 'renesas.git' repo's
'renesas-devel-20180614v2-v4.17' tag. We're adding the R8A77980 PCIe related
device nodes and then enable PCIe on the Condor board. These patches depend
on the R8A77980 PCIe PHY driver support in order to work properly. Note that
in case the PCIe PHY driver is not enabled, the kernel will BUG() due to I/O
space page leak in the PCIe driver...

[1/2] arm64: dts: renesas: r8a77980: add PCIe support
[2/2] arm64: dts: renesas: condor: add PCIe support

WBR, Sergei

[PATCH v2 1/2] arm64: dts: renesas: r8a77980: add PCIe support

[Sergei Shtylyov]

Describe the PCIe PHY, PCIEC, and PCIe bus clock in the R8A77980 device
tree.

Based on the original (and large) patch by Vladimir Barinov.

Signed-off-by: Vladimir Barinov <vladimir.barinov@cogentembedded.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>

---
Changes in version 2:
- merged in the PCIEC patch, renamed the patch, updated the description
  accordingly;
- used R8A77980_PD_ALWAYS_ON in the "power-domains" props
- mentioned Vladimir's original work and added his signoff.

 arch/arm64/boot/dts/renesas/r8a77980.dtsi |   49 ++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

Index: renesas/arch/arm64/boot/dts/renesas/r8a77980.dtsi
===================================================================
--- renesas.orig/arch/arm64/boot/dts/renesas/r8a77980.dtsi
+++ renesas/arch/arm64/boot/dts/renesas/r8a77980.dtsi
@@ -98,6 +98,13 @@
 		clock-frequency = <0>;
 	};
 
+	/* External PCIe clock - can be overridden by the board */
+	pcie_bus_clk: pcie_bus {
+		compatible = "fixed-clock";
+		#clock-cells = <0>;
+		clock-frequency = <0>;
+	};
+
 	psci {
 		compatible = "arm,psci-1.0", "arm,psci-0.2";
 		method = "smc";
@@ -402,6 +409,16 @@
 			status = "disabled";
 		};
 
+		pcie_phy: pcie-phy@e65d0000 {
+			compatible = "renesas,r8a77980-pcie-phy";
+			reg = <0 0xe65d0000 0 0x8000>;
+			#phy-cells = <0>;
+			clocks = <&cpg CPG_MOD 319>;
+			power-domains = <&sysc R8A77980_PD_ALWAYS_ON>;
+			resets = <&cpg 319>;
+			status = "disabled";
+		};
+
 		canfd: can@e66c0000 {
 			compatible = "renesas,r8a77980-canfd",
 				     "renesas,rcar-gen3-canfd";
@@ -653,6 +670,38 @@
 			resets = <&cpg 408>;
 		};
 
+		pciec: pcie@fe000000 {
+			compatible = "renesas,pcie-r8a77980",
+				     "renesas,pcie-rcar-gen3";
+			reg = <0 0xfe000000 0 0x80000>;
+			#address-cells = <3>;
+			#size-cells = <2>;
+			bus-range = <0x00 0xff>;
+			device_type = "pci";
+			ranges = <
+				0x01000000 0 0x00000000 0 0xfe100000 0 0x0100000
+				0x02000000 0 0xfe200000 0 0xfe200000 0 0x0200000
+				0x02000000 0 0x30000000 0 0x30000000 0 0x8000000
+				0x42000000 0 0x38000000 0 0x38000000 0 0x8000000
+			>;
+			dma-ranges = <0x42000000 0 0x40000000 0 0x40000000
+				      0 0x80000000>;
+			interrupts = <GIC_SPI 148 IRQ_TYPE_LEVEL_HIGH>,
+				     <GIC_SPI 149 IRQ_TYPE_LEVEL_HIGH>,
+				     <GIC_SPI 150 IRQ_TYPE_LEVEL_HIGH>;
+			#interrupt-cells = <1>;
+			interrupt-map-mask = <0 0 0 0>;
+			interrupt-map = <0 0 0 0 &gic GIC_SPI 148
+					 IRQ_TYPE_LEVEL_HIGH>;
+			clocks = <&cpg CPG_MOD 319>, <&pcie_bus_clk>;
+			clock-names = "pcie", "pcie_bus";
+			power-domains = <&sysc R8A77980_PD_ALWAYS_ON>;
+			resets = <&cpg 319>;
+			phys = <&pcie_phy>;
+			phy-names = "pcie";
+			status = "disabled";
+		};
+
 		vspd0: vsp@fea20000 {
 			compatible = "renesas,vsp2";
 			reg = <0 0xfea20000 0 0x5000>;

Re: [PATCH v2 1/2] arm64: dts: renesas: r8a77980: add PCIe support

[Simon Horman]

On Thu, Jun 14, 2018 at 10:17:46PM +0300, Sergei Shtylyov wrote:
> Describe the PCIe PHY, PCIEC, and PCIe bus clock in the R8A77980 device
> tree.
> 
> Based on the original (and large) patch by Vladimir Barinov.
> 
> Signed-off-by: Vladimir Barinov <vladimir.barinov@cogentembedded.com>
> Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>

This looks good to me.

Reviewed-by: Simon Horman <horms+renesas@verge.net.au>

I am marking it as deferred pending acceptance of the
renesas,r8a77980-pcie-phy binding. Please repost or otherwise
ping me once that has happened.

[PATCH v2 2/2] arm64: dts: renesas: condor: add PCIe support

[Sergei Shtylyov]

Enable PCIe PHY and PCIEC and specify the PCIe bus clock for the Condor
board.

Based on the original (and large) patch by Vladimir Barinov.

Signed-off-by: Vladimir Barinov <vladimir.barinov@cogentembedded.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>

---
Changes in version 2:
- mentioned Vladimir's original work and added his signoff;
- refreshed the patch.

 arch/arm64/boot/dts/renesas/r8a77980-condor.dts |   12 ++++++++++++
 1 file changed, 12 insertions(+)

Index: renesas/arch/arm64/boot/dts/renesas/r8a77980-condor.dts
===================================================================
--- renesas.orig/arch/arm64/boot/dts/renesas/r8a77980-condor.dts
+++ renesas/arch/arm64/boot/dts/renesas/r8a77980-condor.dts
@@ -223,6 +223,18 @@
 	status = "okay";
 };
 
+&pciec {
+	status = "okay";
+};
+
+&pcie_bus_clk {
+	clock-frequency = <100000000>;
+};
+
+&pcie_phy {
+	status = "okay";
+};
+
 &pfc {
 	avb_pins: avb {
 		groups = "avb_mdio", "avb_rgmii";

Re: [PATCH v2 0/2] Add R8A77980/Condor PCIe support

[Simon Horman]

On Thu, Jun 14, 2018 at 10:16:32PM +0300, Sergei Shtylyov wrote:
> Hello!
> 
> Here's the set of 2 patches against Simon Horman's 'renesas.git' repo's
> 'renesas-devel-20180614v2-v4.17' tag. We're adding the R8A77980 PCIe related
> device nodes and then enable PCIe on the Condor board. These patches depend
> on the R8A77980 PCIe PHY driver support in order to work properly. Note that
> in case the PCIe PHY driver is not enabled, the kernel will BUG() due to I/O
> space page leak in the PCIe driver...

Is that problem specific to the presence of PCIe nodes for
condor/r8a77980 or is it also true of other (R-Car) boards where
PCIe is enabled?

Regardless, it sounds like these patches expose a kernel bug.
Is it being fixed?

> 
> [1/2] arm64: dts: renesas: r8a77980: add PCIe support
> [2/2] arm64: dts: renesas: condor: add PCIe support
> 
> WBR, Sergei
> 

Re: [PATCH v2 0/2] Add R8A77980/Condor PCIe support

[Sergei Shtylyov]

On 06/18/2018 12:36 PM, Simon Horman wrote:

>> Here's the set of 2 patches against Simon Horman's 'renesas.git' repo's
>> 'renesas-devel-20180614v2-v4.17' tag. We're adding the R8A77980 PCIe related
>> device nodes and then enable PCIe on the Condor board. These patches depend
>> on the R8A77980 PCIe PHY driver support in order to work properly. Note that
>> in case the PCIe PHY driver is not enabled, the kernel will BUG() due to I/O
>> space page leak in the PCIe driver...
> 
> Is that problem specific to the presence of PCIe nodes for
> condor/r8a77980

   The nodes are safe unless they are enabled, so the Condor patch may be
deferred untl I fix the PCI code.
> condor/r8a77980 or is it also true of other (R-Car) boards where
> PCIe is enabled?

   The leak happens every time the driver fails to probe later than
pci_remap_iospace() is called but the BUG_ON() is only triggered by rhe 2nd try
with EPROBE_DEFER returned previously.

> Regardless, it sounds like these patches expose a kernel bug.
> Is it being fixed?

   I'm working on a fix (which embraces several PCI drivers)...

>> [1/2] arm64: dts: renesas: r8a77980: add PCIe support
>> [2/2] arm64: dts: renesas: condor: add PCIe support

WBR, Sergei

Re: [PATCH v2 0/2] Add R8A77980/Condor PCIe support

[Simon Horman]

On Mon, Jun 18, 2018 at 07:16:47PM +0300, Sergei Shtylyov wrote:
> On 06/18/2018 12:36 PM, Simon Horman wrote:
> 
> >> Here's the set of 2 patches against Simon Horman's 'renesas.git' repo's
> >> 'renesas-devel-20180614v2-v4.17' tag. We're adding the R8A77980 PCIe related
> >> device nodes and then enable PCIe on the Condor board. These patches depend
> >> on the R8A77980 PCIe PHY driver support in order to work properly. Note that
> >> in case the PCIe PHY driver is not enabled, the kernel will BUG() due to I/O
> >> space page leak in the PCIe driver...
> > 
> > Is that problem specific to the presence of PCIe nodes for
> > condor/r8a77980
> 
>    The nodes are safe unless they are enabled, so the Condor patch may be
> deferred untl I fix the PCI code.

Understood. I would expect that given this is a but a fix
would be backported to -stable in due course. But I agree its
best to be cautious here.

> > condor/r8a77980 or is it also true of other (R-Car) boards where
> > PCIe is enabled?
> 
>    The leak happens every time the driver fails to probe later than
> pci_remap_iospace() is called but the BUG_ON() is only triggered by rhe 2nd try
> with EPROBE_DEFER returned previously.
> 
> > Regardless, it sounds like these patches expose a kernel bug.
> > Is it being fixed?
> 
>    I'm working on a fix (which embraces several PCI drivers)...
> 
> >> [1/2] arm64: dts: renesas: r8a77980: add PCIe support
> >> [2/2] arm64: dts: renesas: condor: add PCIe support
> 
> WBR, Sergei
>