Skip to contentSkip to conversation

arm64: tegra: Fix Tegra234 PCIe interrupt-map

PATCH Superseded Brad Griffis 5 participants 8 messages Dec 12, 2024 – Dec 14, 2024

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm64: tegra: Fix Tegra234 PCIe interrupt-map

For interrupt-map entries, the DTS specification requires that #address-cells is defined for both the child node and the interrupt parent. For the PCIe interrupt-map entries, the parent node ("gic") has not specified #address-cells. The existing layout of the PCIe interrupt-map entries indicates that it assumes that #address-cells is zero for this node.

Explicitly set #address-cells to zero for "gic" so that it complies with the device tree specification.

---
 arch/arm64/boot/dts/nvidia/tegra234.dtsi | 1 +
 1 file changed, 1 insertion(+)
arch/arm64/boot/dts/nvidia/tegra234.dtsi +1 -0
index 984c85eab41a..e1c07c99e9bd 100644
@@ -4010,6 +4010,7 @@ ccplex@e000000 {
 
 		gic: interrupt-controller@f400000 {
 			compatible = "arm,gic-v3";
+			#address-cells = <0>;
 			reg = <0x0 0x0f400000 0x0 0x010000>, /* GICD */
 			      <0x0 0x0f440000 0x0 0x200000>; /* GICR */
 			interrupt-parent = <&gic>;
Signature · 1 line
2.34.1
maintainer
#1 lore raw
Quoted 20 lines · Brad Griffis

For interrupt-map entries, the DTS specification requires that #address-cells is defined for both the child node and the interrupt parent. For the PCIe interrupt-map entries, the parent node ("gic") has not specified #address-cells. The existing layout of the PCIe interrupt-map entries indicates that it assumes that #address-cells is zero for this node.

Explicitly set #address-cells to zero for "gic" so that it complies with the device tree specification.

---
  arch/arm64/boot/dts/nvidia/tegra234.dtsi | 1 +
  1 file changed, 1 insertion(+)
arch/arm64/boot/dts/nvidia/tegra234.dtsi +0 -0
index 984c85eab41a..e1c07c99e9bd 100644

arch/arm64/boot/dts/nvidia/tegra234.dtsi ccplex@e000000 { · line 4016

Quoted 7 lines
arch/arm64/boot/dts/nvidia/tegra234.dtsi +1 -0
@@ -4010,6 +4010,7 @@ ccplex@e000000 {
 
  		gic: interrupt-controller@f400000 {
  			compatible = "arm,gic-v3";
+			#address-cells = <0>;
  			reg = <0x0 0x0f400000 0x0 0x010000>, /* GICD */
  			      <0x0 0x0f440000 0x0 0x200000>; /* GICR */
  			interrupt-parent = <&gic>;

Can you add the fixes tag for this?

BTW, I had a quick look and this appears to be a common problem across all Tegra devices, Tegra210, Tegra186 and Tegra194. Should we fix up all while we are at it?

Thanks
Jon

Signature · 1 line
nvpublic
#2 lore raw
Quoted 32 lines · Jon Hunter
Quoted 20 lines · Brad Griffis

For interrupt-map entries, the DTS specification requires that #address-cells is defined for both the child node and the interrupt parent. For the PCIe interrupt-map entries, the parent node ("gic") has not specified #address-cells. The existing layout of the PCIe interrupt-map entries indicates that it assumes that #address-cells is zero for this node.

Explicitly set #address-cells to zero for "gic" so that it complies with the device tree specification.

---
  arch/arm64/boot/dts/nvidia/tegra234.dtsi | 1 +
  1 file changed, 1 insertion(+)
arch/arm64/boot/dts/nvidia/tegra234.dtsi +0 -0
index 984c85eab41a..e1c07c99e9bd 100644

arch/arm64/boot/dts/nvidia/tegra234.dtsi ccplex@e000000 { · line 4015

Quoted 6 lines
arch/arm64/boot/dts/nvidia/tegra234.dtsi +1 -0
@@ -4010,6 +4010,7 @@ ccplex@e000000 {
  		gic: interrupt-controller@f400000 {
  			compatible = "arm,gic-v3";
+			#address-cells = <0>;
  			reg = <0x0 0x0f400000 0x0 0x010000>, /* GICD */
  			      <0x0 0x0f440000 0x0 0x200000>; /* GICR */
  			interrupt-parent = <&gic>;

Can you add the fixes tag for this?

Quoted 3 lines

BTW, I had a quick look and this appears to be a common problem across all Tegra devices, Tegra210, Tegra186 and Tegra194. Should we fix up all while we are at it?

Yeah, I think we should fix these up going as far back as Tegra20. I don't think we've ever seen any fallout from this being broken and I don't exactly understand why. I'll dig through some code and run some tests to see if I can get to the bottom of this.

I can take care of fixing up the older boards.

Thierry

#3 lore raw
In reply to Brad Griffis
Quoted 7 lines · Brad Griffis

For interrupt-map entries, the DTS specification requires that #address-cells is defined for both the child node and the interrupt parent. For the PCIe interrupt-map entries, the parent node ("gic") has not specified #address-cells. The existing layout of the PCIe interrupt-map entries indicates that it assumes that #address-cells is zero for this node.

Quoted 5 lines

Explicitly set #address-cells to zero for "gic" so that it complies with the device tree specification.

---

So entire nvidia works on old kernel and sends patches rebased on old kernel? Please start working on mainline.

I give the same feedback multiple times and still nvidia repeats the same mistakes.

You CC-ed an address, which suggests you do not work on mainline kernel or you do not use get_maintainers.pl/b4/patman. Please rebase and always work on mainline or start using mentioned tools, so correct addresses will be used.

Best regards,
Krzysztof

maintainer
#4 lore raw

Hi Krzysztof!

Quoted 8 lines · Krzysztof Kozlowski
Quoted 7 lines · Brad Griffis

For interrupt-map entries, the DTS specification requires that #address-cells is defined for both the child node and the interrupt parent. For the PCIe interrupt-map entries, the parent node ("gic") has not specified #address-cells. The existing layout of the PCIe interrupt-map entries indicates that it assumes that #address-cells is zero for this node.

Quoted 8 lines
Quoted 5 lines

Explicitly set #address-cells to zero for "gic" so that it complies with the device tree specification.

---

So entire nvidia works on old kernel and sends patches rebased on old kernel? Please start working on mainline.

You know that is not the case! Yes there may be a few cases of people using your old email, but not all of us :-)

Quoted 7 lines

I give the same feedback multiple times and still nvidia repeats the same mistakes.

You CC-ed an address, which suggests you do not work on mainline kernel or you do not use get_maintainers.pl/b4/patman. Please rebase and always work on mainline or start using mentioned tools, so correct addresses will be used.

This applies on top of today's -next and so I don't see any need to resend.

FYI, Brad has been working improve our testing for mainline and -next and so I know he is working on mainline.

So are we perfect? No. But we are working on getting better!

Cheers
Jon

Signature · 1 line
nvpublic
#5 lore raw
Quoted 17 lines · Jon Hunter

Hi Krzysztof!

Quoted 14 lines · Krzysztof Kozlowski
Quoted 12 lines · Brad Griffis

For interrupt-map entries, the DTS specification requires that #address-cells is defined for both the child node and the interrupt parent. For the PCIe interrupt-map entries, the parent node ("gic") has not specified #address-cells. The existing layout of the PCIe interrupt-map entries indicates that it assumes that #address-cells is zero for this node.

Explicitly set #address-cells to zero for "gic" so that it complies with the device tree specification.

---

Quoted 6 lines
Quoted 2 lines

So entire nvidia works on old kernel and sends patches rebased on old kernel? Please start working on mainline.

You know that is not the case! Yes there may be a few cases of people using your old email, but not all of us :-)

Yeah, that was just an assumption, but three independent people are Ccing address which does not exist in the kernel and it is impossible to get/deduce/invent. Three patchsets from three different people...

Best regards,
Krzysztof

#6 lore raw
Quoted 6 lines · Krzysztof Kozlowski

Yeah, that was just an assumption, but three independent people are Ccing address which does not exist in the kernel and it is impossible to get/deduce/invent. Three patchsets from three different people...

Best regards,
Krzysztof

I am developing and testing on mainline. However, when I checked the get_maintainer.pl output on first glance it looked unchanged and I copied/pasted an old "git send-email" command that had the older email address.

And you are probably getting multiple of them due to me "helping" other people by sharing some of the commands/syntax.

I apologize for the mistake. I will check that more carefully from now on. I sent out v2 and made sure to get this correct!

#7 lore raw
Quoted 5 lines · Brad Griffis
Quoted 4 lines · Krzysztof Kozlowski

Yeah, that was just an assumption, but three independent people are Ccing address which does not exist in the kernel and it is impossible to get/deduce/invent. Three patchsets from three different people...

Quoted 7 lines
Quoted 2 lines

Best regards,
Krzysztof

I am developing and testing on mainline. However, when I checked the get_maintainer.pl output on first glance it looked unchanged and I copied/pasted an old "git send-email" command that had the older email address.

You should not "cache" the output of get_maintainers.pl EVER. Addresses change, people are added or removed. Just use b4 or simple scripts/aliases (like Rob's send email identity or
https://github.com/krzk/tools/blob/master/linux/.bash_aliases_linux#L92 ).

Best regards,
Krzysztof