Skip to contentSkip to conversation

ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

PATCH New Fabio Estevam 3 participants 8 messages Feb 12, 2014 – Feb 13, 2014

Subsystems (2)
lore
Share & tools
lore

Fabio Estevam

author reviewer
#0 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

From: Fabio Estevam <[email protected]>

According to Documentation/devicetree/bindings/regulator/regulator.txt regulator nodes should not be placed under 'simple-bus'.

Mark Rutland also explains about it at:
http://www.spinics.net/lists/linux-usb/msg101497.html

---
Shawn,

I can convert other dts files if you are fine with this approach.

 arch/arm/boot/dts/imx6qdl-sabresd.dtsi | 51 ++++++++++++++--------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +21 -30
index 0d816d3..d7df5b2 100644
@@ -18,38 +18,29 @@
 		reg = <0x10000000 0x40000000>;
 	};
 
-	regulators {
-		compatible = "simple-bus";
-		#address-cells = <1>;
-		#size-cells = <0>;
-
-		reg_usb_otg_vbus: regulator at 0 {
-			compatible = "regulator-fixed";
-			reg = <0>;
-			regulator-name = "usb_otg_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio3 22 0>;
-			enable-active-high;
-		};
+	reg_usb_otg_vbus: regulator at 0 {
+		compatible = "regulator-fixed";
+		regulator-name = "usb_otg_vbus";
+		regulator-min-microvolt = <5000000>;
+		regulator-max-microvolt = <5000000>;
+		gpio = <&gpio3 22 0>;
+		enable-active-high;
+	};
 
-		reg_usb_h1_vbus: regulator at 1 {
-			compatible = "regulator-fixed";
-			reg = <1>;
-			regulator-name = "usb_h1_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio1 29 0>;
-			enable-active-high;
-		};
+	reg_usb_h1_vbus: regulator at 1 {
+		compatible = "regulator-fixed";
+		regulator-name = "usb_h1_vbus";
+		regulator-min-microvolt = <5000000>;
+		regulator-max-microvolt = <5000000>;
+		gpio = <&gpio1 29 0>;
+		enable-active-high;
+	};
 
-		reg_audio: regulator at 2 {
-			compatible = "regulator-fixed";
-			reg = <2>;
-			regulator-name = "wm8962-supply";
-			gpio = <&gpio4 10 0>;
-			enable-active-high;
-		};
+	reg_audio: regulator at 2 {
+		compatible = "regulator-fixed";
+		regulator-name = "wm8962-supply";
+		gpio = <&gpio4 10 0>;
+		enable-active-high;
 	};
 
 	gpio-keys {
Signature · 1 line
1.8.1.2
#1 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 4 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

According to Documentation/devicetree/bindings/regulator/regulator.txt regulator nodes should not be placed under 'simple-bus'.

I failed to read that statement from the binding doc. Can you quote the doc specifically on that statement?

Quoted 9 lines

Mark Rutland also explains about it at:
http://www.spinics.net/lists/linux-usb/msg101497.html

---
Shawn,

I can convert other dts files if you are fine with this approach.

I take it as an unnecessary churn, unless I see the consensus from most of DT maintainers and arm-soc folks that we should make this change. And see comment below ...

Quoted 17 lines
 arch/arm/boot/dts/imx6qdl-sabresd.dtsi | 51 ++++++++++++++--------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +0 -5
index 0d816d3..d7df5b2 100644
@@ -18,38 +18,29 @@
 		reg = <0x10000000 0x40000000>;
 	};
 
-	regulators {
-		compatible = "simple-bus";
-		#address-cells = <1>;
-		#size-cells = <0>;
-

arch/arm/boot/dts/imx6qdl-sabresd.dtsi line 21

Quoted 10 lines
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +1 -9
-		reg_usb_otg_vbus: regulator at 0 {
-			compatible = "regulator-fixed";
-			reg = <0>;
-			regulator-name = "usb_otg_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio3 22 0>;
-			enable-active-high;
-		};
+	reg_usb_otg_vbus: regulator at 0 {

nodename at num should only be used for nodes that have 'reg' property. Why are you dropping 'reg' property here? Right, it does not compile if you do not drop it. You can take it as a reason of why I endorse simple-bus regulators container.

Shawn

Quoted 44 lines
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +20 -16
+		compatible = "regulator-fixed";
+		regulator-name = "usb_otg_vbus";
+		regulator-min-microvolt = <5000000>;
+		regulator-max-microvolt = <5000000>;
+		gpio = <&gpio3 22 0>;
+		enable-active-high;
+	};
 
-		reg_usb_h1_vbus: regulator at 1 {
-			compatible = "regulator-fixed";
-			reg = <1>;
-			regulator-name = "usb_h1_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio1 29 0>;
-			enable-active-high;
-		};
+	reg_usb_h1_vbus: regulator at 1 {
+		compatible = "regulator-fixed";
+		regulator-name = "usb_h1_vbus";
+		regulator-min-microvolt = <5000000>;
+		regulator-max-microvolt = <5000000>;
+		gpio = <&gpio1 29 0>;
+		enable-active-high;
+	};
 
-		reg_audio: regulator at 2 {
-			compatible = "regulator-fixed";
-			reg = <2>;
-			regulator-name = "wm8962-supply";
-			gpio = <&gpio4 10 0>;
-			enable-active-high;
-		};
+	reg_audio: regulator at 2 {
+		compatible = "regulator-fixed";
+		regulator-name = "wm8962-supply";
+		gpio = <&gpio4 10 0>;
+		enable-active-high;
 	};
 
 	gpio-keys {
Signature · 1 line
1.8.1.2
#2 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 8 lines · Shawn Guo
Quoted 4 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

According to Documentation/devicetree/bindings/regulator/regulator.txt regulator nodes should not be placed under 'simple-bus'.

I failed to read that statement from the binding doc. Can you quote the doc specifically on that statement?

The statement is slightly wrong. Rather, regulators are not required to be on a simple bus.

In this particular case the simple-bus is being used incorrectly. I describe some of this in the link Fabio provided immediately below, and I have further comments specific to this case.

Quoted 9 lines
Quoted 9 lines

Mark Rutland also explains about it at:
http://www.spinics.net/lists/linux-usb/msg101497.html

---
Shawn,

Quoted 5 lines
Quoted 1 line

I can convert other dts files if you are fine with this approach.

I take it as an unnecessary churn, unless I see the consensus from most of DT maintainers and arm-soc folks that we should make this change. And see comment below ...

My concern is that the simple-bus is being used in a nonsensical way, with a meaningless address space and reg values on children. While this currently works it is not correct, and it's not even necessary. I would like to limit the misuse of these constructs so as to prevent others from making the same mistakes.

Quoted 25 lines
Quoted 25 lines

 arch/arm/boot/dts/imx6qdl-sabresd.dtsi | 51 ++++++++++++++--------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +0 -12
index 0d816d3..d7df5b2 100644
@@ -18,38 +18,29 @@
 		reg = <0x10000000 0x40000000>;
 	};
 
-	regulators {
-		compatible = "simple-bus";
-		#address-cells = <1>;
-		#size-cells = <0>;
-
-		reg_usb_otg_vbus: regulator at 0 {
-			compatible = "regulator-fixed";
-			reg = <0>;
-			regulator-name = "usb_otg_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio3 22 0>;
Quoted 8 lines

arch/arm/boot/dts/imx6qdl-sabresd.dtsi line 21

Quoted 3 lines
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +1 -2
-			enable-active-high;
-		};
+	reg_usb_otg_vbus: regulator at 0 {

nodename at num should only be used for nodes that have 'reg' property. Why are you dropping 'reg' property here? Right, it does not compile if you do not drop it. You can take it as a reason of why I endorse simple-bus regulators container.

I don't think the logical jump from "it does not compile" to "I endorse simple-bus regulators container" makes any sense. They're two separate issues.

The unit-addresses and reg values can be dropped. They are not in the regulator-fixed binding, and were never necessary. All that's necessary is a unique name, and unit-addresses (which required a reg) were misused to provide uniqueness. With that fixed, this will compile.

The question to ask is what address space does the simple-bus represent? The regulators never defined one in their binding, and the simple-bus is missing a _required_ ranges property, so it's not in the MMIO space of the parent. Yet somehow it has a single address cell and no size cells.

This is _clearly_ not a simple-bus. Either it needs to map to the parent address space in some way (which would mean dropping the reg values and unit addresses as they would be clearly wrong), or it needs to be replaced by something that isn't a simple-bus. For the latter case that would mean you couldn't have MMIO regulators in the same container as fixed regulators, which would be a bit useless.

As far as I can see, a regulator container node is pointless. It describes no topological information yet pretends to.

Mark.

Quoted 49 lines

Shawn

Quoted 44 lines
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +20 -16
+		compatible = "regulator-fixed";
+		regulator-name = "usb_otg_vbus";
+		regulator-min-microvolt = <5000000>;
+		regulator-max-microvolt = <5000000>;
+		gpio = <&gpio3 22 0>;
+		enable-active-high;
+	};
 
-		reg_usb_h1_vbus: regulator at 1 {
-			compatible = "regulator-fixed";
-			reg = <1>;
-			regulator-name = "usb_h1_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio1 29 0>;
-			enable-active-high;
-		};
+	reg_usb_h1_vbus: regulator at 1 {
+		compatible = "regulator-fixed";
+		regulator-name = "usb_h1_vbus";
+		regulator-min-microvolt = <5000000>;
+		regulator-max-microvolt = <5000000>;
+		gpio = <&gpio1 29 0>;
+		enable-active-high;
+	};
 
-		reg_audio: regulator at 2 {
-			compatible = "regulator-fixed";
-			reg = <2>;
-			regulator-name = "wm8962-supply";
-			gpio = <&gpio4 10 0>;
-			enable-active-high;
-		};
+	reg_audio: regulator at 2 {
+		compatible = "regulator-fixed";
+		regulator-name = "wm8962-supply";
+		gpio = <&gpio4 10 0>;
+		enable-active-high;
 	};
 
 	gpio-keys {
Signature · 1 line
1.8.1.2
#3 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 9 lines · Mark Rutland
Quoted 3 lines

I take it as an unnecessary churn, unless I see the consensus from most of DT maintainers and arm-soc folks that we should make this change. And see comment below ...

My concern is that the simple-bus is being used in a nonsensical way, with a meaningless address space and reg values on children. While this currently works it is not correct, and it's not even necessary. I would like to limit the misuse of these constructs so as to prevent others from making the same mistakes.

Unfortunately, it's already been used quite widely. This is not IMX specific, and you can grep arch/arm/boot/dts to get the idea. I'm not sure if it's worth the churn to 'fix' it. In these days, arm-soc folks are quite sensitive to the number of IMX patches and change sets. 50 LOC change for one board dts, and we now have ~70 files for IMX. That's why I would like to get the agreement from arm-soc folks that we should really make such change.

Quoted 35 lines
Quoted 26 lines
Quoted 26 lines


 arch/arm/boot/dts/imx6qdl-sabresd.dtsi | 51 ++++++++++++++--------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +0 -12
index 0d816d3..d7df5b2 100644
@@ -18,38 +18,29 @@
 		reg = <0x10000000 0x40000000>;
 	};
 
-	regulators {
-		compatible = "simple-bus";
-		#address-cells = <1>;
-		#size-cells = <0>;
-
-		reg_usb_otg_vbus: regulator at 0 {
-			compatible = "regulator-fixed";
-			reg = <0>;
-			regulator-name = "usb_otg_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio3 22 0>;
Quoted 8 lines

arch/arm/boot/dts/imx6qdl-sabresd.dtsi line 21

Quoted 3 lines
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +1 -2
-			enable-active-high;
-		};
+	reg_usb_otg_vbus: regulator at 0 {

nodename at num should only be used for nodes that have 'reg' property. Why are you dropping 'reg' property here? Right, it does not compile if you do not drop it. You can take it as a reason of why I endorse simple-bus regulators container.

Quoted 3 lines

I don't think the logical jump from "it does not compile" to "I endorse simple-bus regulators container" makes any sense. They're two separate issues.

You're right :)

Quoted 5 lines

The unit-addresses and reg values can be dropped. They are not in the regulator-fixed binding, and were never necessary. All that's necessary is a unique name, and unit-addresses (which required a reg) were misused to provide uniqueness. With that fixed, this will compile.

IIRC, ePAPR recommends to use generic node name and have unit-addresses for uniqueness. That's why I sent patch [1] to follow the way that tegra and other platforms names them. Now you're suggesting to go back?

Shawn

[1] http://www.spinics.net/lists/arm-kernel/msg284474.html

#4 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 11 lines · Shawn Guo
Quoted 9 lines · Mark Rutland
Quoted 3 lines

I take it as an unnecessary churn, unless I see the consensus from most of DT maintainers and arm-soc folks that we should make this change. And see comment below ...

My concern is that the simple-bus is being used in a nonsensical way, with a meaningless address space and reg values on children. While this currently works it is not correct, and it's not even necessary. I would like to limit the misuse of these constructs so as to prevent others from making the same mistakes.

Quoted 7 lines

Unfortunately, it's already been used quite widely. This is not IMX specific, and you can grep arch/arm/boot/dts to get the idea. I'm not sure if it's worth the churn to 'fix' it. In these days, arm-soc folks are quite sensitive to the number of IMX patches and change sets. 50 LOC change for one board dts, and we now have ~70 files for IMX. That's why I would like to get the agreement from arm-soc folks that we should really make such change.

As it stands, the dts are buggy. I can appreciate that you don't feel this is important, but I do. It's not just an IMX issue, there is widespread misunderstanding and abuse of simple-bus.

Said abuse is relying on current Linux implementation details, and that can and will create problems if and when probing code is changed. There's no good reason for abusing the binding when we can get it right today.

We don't necessarily have to get rid of every simple bus in use as a regulator container. While I don't like that use, I am happy to accept them as long as they are used correctly, with a valid ranges property and with children having sensible reg values (or no reg values if the children don't need them as part of their binding).

Quoted 48 lines
Quoted 36 lines
Quoted 27 lines
Quoted 27 lines



 arch/arm/boot/dts/imx6qdl-sabresd.dtsi | 51 ++++++++++++++--------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +0 -12
index 0d816d3..d7df5b2 100644
@@ -18,38 +18,29 @@
 		reg = <0x10000000 0x40000000>;
 	};
 
-	regulators {
-		compatible = "simple-bus";
-		#address-cells = <1>;
-		#size-cells = <0>;
-
-		reg_usb_otg_vbus: regulator at 0 {
-			compatible = "regulator-fixed";
-			reg = <0>;
-			regulator-name = "usb_otg_vbus";
-			regulator-min-microvolt = <5000000>;
-			regulator-max-microvolt = <5000000>;
-			gpio = <&gpio3 22 0>;
Quoted 8 lines

arch/arm/boot/dts/imx6qdl-sabresd.dtsi line 21

Quoted 3 lines
arch/arm/boot/dts/imx6qdl-sabresd.dtsi +1 -2
-			enable-active-high;
-		};
+	reg_usb_otg_vbus: regulator at 0 {

nodename at num should only be used for nodes that have 'reg' property. Why are you dropping 'reg' property here? Right, it does not compile if you do not drop it. You can take it as a reason of why I endorse simple-bus regulators container.

Quoted 3 lines

I don't think the logical jump from "it does not compile" to "I endorse simple-bus regulators container" makes any sense. They're two separate issues.

You're right :)

Quoted 5 lines

The unit-addresses and reg values can be dropped. They are not in the regulator-fixed binding, and were never necessary. All that's necessary is a unique name, and unit-addresses (which required a reg) were misused to provide uniqueness. With that fixed, this will compile.

Quoted 4 lines

IIRC, ePAPR recommends to use generic node name and have unit-addresses for uniqueness. That's why I sent patch [1] to follow the way that tegra and other platforms names them. Now you're suggesting to go back?

As far as I can see, that's not what ePAPR says:

If the node has no reg property, the @ and unit-address must be omitted and the node-name alone differentiates the node from other nodes at the same level in the tree.

If a binding doesn't require a reg property, then the node shouldn't have a reg property. If that's true then the node should not have a unit-address.

Using generic names can certainly help to make dts more legible, and we don't have to go against that. It's just as easy to understand regulator_1 and regulator at 1.

I can appreciate that this area may be unclear, and that fixing this up creates churn for what you feel is a minor issue. However, I don't see that as a reason to perpetuate wrongness.

Thanks,
Mark.

Quoted 10 lines

Shawn

[1] http://www.spinics.net/lists/arm-kernel/msg284474.html

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

#5 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 6 lines · Mark Rutland

As it stands, the dts are buggy. I can appreciate that you don't feel this is important, but I do. It's not just an IMX issue, there is widespread misunderstanding and abuse of simple-bus.

Said abuse is relying on current Linux implementation details, and that can and will create problems if and when probing code is changed.

The reality is the code already gets no chance to change on this regard, considering the requirement that we need to maintain the interface between kernel and DT as ABI. The dts have been there like this for 10 kernel releases or so.

Quoted 2 lines

There's no good reason for abusing the binding when we can get it right today.

As I said from the beginning, though I personally take it as a churn, I will still be happy take the change, as long as my upstream folks are also fine with it.

Shawn

author reviewer
#6 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 5 lines · Shawn Guo
Quoted 4 lines · Mark Rutland

As it stands, the dts are buggy. I can appreciate that you don't feel this is important, but I do. It's not just an IMX issue, there is widespread misunderstanding and abuse of simple-bus.

Quoted 7 lines
Quoted 2 lines

Said abuse is relying on current Linux implementation details, and that can and will create problems if and when probing code is changed.

The reality is the code already gets no chance to change on this regard, considering the requirement that we need to maintain the interface between kernel and DT as ABI. The dts have been there like this for 10 kernel releases or so.

What breakage do you see with this patch?

#7 lore raw

PATCH 1/1 ARM: dts: imx6qdl-sabresd: Do not place regulator nodes under simple-bus

Quoted 9 lines · Fabio Estevam
Quoted 8 lines · Shawn Guo
Quoted 6 lines · Mark Rutland

As it stands, the dts are buggy. I can appreciate that you don't feel this is important, but I do. It's not just an IMX issue, there is widespread misunderstanding and abuse of simple-bus.

Said abuse is relying on current Linux implementation details, and that can and will create problems if and when probing code is changed.

Quoted 6 lines
Quoted 4 lines

The reality is the code already gets no chance to change on this regard, considering the requirement that we need to maintain the interface between kernel and DT as ABI. The dts have been there like this for 10 kernel releases or so.

What breakage do you see with this patch?

I'm not talking about this patch. I'm replying to Mark's comment saying some day the kernel probing code is changed.

Shawn