Skip to contentSkip to conversation

ARM: dts: imx6qdl: make pinctrl nodes board specific

RFC PATCH New Shawn Guo 5 participants 12 messages Oct 25, 2013 – Nov 5, 2013

Subsystems (2)
lore
Share & tools
lore

Shawn Guo

author
#0 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Currently, all pinctrl setting nodes are defined in <soc>.dtsi, so that boards that share the same pinctrl setting do not have to define it time and time again in <board>.dts. However, along with the devices and use cases being added continuously, the pinctrl setting nodes under iomuxc becomes more than expected. This bloats device tree blob for particular board unnecessarily since only a small subset of those pinctrl setting nodes will be used by the board. It impacts not only the DTB file size but also the run-time device tree lookup efficiency.

The patch proposes a solution to avoid this device tree bloating problem while still keeping boards share the common pinctrl setting data by using DTC macro support. It creates imx6qdl-pingrp.h and move all those pinctrl setting data into there as macro definitions. The <board>.dts will instead define the pinctrl setting nodes that are necessary for the board by referring to the macros in imx6qdl-pingrp.h, so that only the pinctrl setting data that will be used by the board will get compiled into the DTB for the board.

With the changes, the pinctrl setting nodes becomes local to particular board, and it makes no sense to continue numbering the setting for given peripheral. Thus, all the pinctrl phandler name gets updated to have only peripheral name in there.

Using imx6q-sabresd.dtb today as example, the change shrinks the file from 37KiB to 26KiB.

---
Russell,

With this approach, I will have no concern with adding another full group of USDHC pins with only one different from the existing group.

#define MX6QDL_USDHC1_PINGRP1_DAT3CD \
        MX6QDL_PAD_SD1_CMD__SD1_CMD    0x17059 \
        MX6QDL_PAD_SD1_CLK__SD1_CLK    0x10059 \
        MX6QDL_PAD_SD1_DAT0__SD1_DATA0 0x17059 \
        MX6QDL_PAD_SD1_DAT1__SD1_DATA1 0x17059 \
        MX6QDL_PAD_SD1_DAT2__SD1_DATA2 0x17059 \
        MX6QDL_PAD_SD1_DAT3__SD1_DATA3 0x13059 \
        MX6QDL_PAD_NANDF_D0__SD1_DATA4 0x17059 \
        MX6QDL_PAD_NANDF_D1__SD1_DATA5 0x17059 \
        MX6QDL_PAD_NANDF_D2__SD1_DATA6 0x17059 \
        MX6QDL_PAD_NANDF_D3__SD1_DATA7 0x17059

Shawn

 arch/arm/boot/dts/imx6dl.dtsi              |    1 +
 arch/arm/boot/dts/imx6q-arm2.dts           |   44 +-
 arch/arm/boot/dts/imx6q-phytec-pfla02.dtsi |   44 +-
 arch/arm/boot/dts/imx6q-sabrelite.dts      |   50 +-
 arch/arm/boot/dts/imx6q-sbc6x.dts          |   29 +-
 arch/arm/boot/dts/imx6q-udoo.dts           |   16 +-
 arch/arm/boot/dts/imx6q.dtsi               |    1 +
 arch/arm/boot/dts/imx6qdl-pingrp.h         |  503 +++++++++++++++++++
 arch/arm/boot/dts/imx6qdl-sabreauto.dtsi   |   58 ++-
 arch/arm/boot/dts/imx6qdl-sabresd.dtsi     |   62 ++-
 arch/arm/boot/dts/imx6qdl-wandboard.dtsi   |   62 ++-
 arch/arm/boot/dts/imx6qdl.dtsi             |  738 ----------------------------
 12 files changed, 799 insertions(+), 809 deletions(-)
 create mode 100644 arch/arm/boot/dts/imx6qdl-pingrp.h
arch/arm/boot/dts/imx6dl.dtsi +1 -0
index 9e8ae11..4d9189f 100644
@@ -9,6 +9,7 @@
  */
 
 #include "imx6dl-pinfunc.h"
+#include "imx6qdl-pingrp.h"
 #include "imx6qdl.dtsi"
 
 / {
arch/arm/boot/dts/imx6q-arm2.dts +33 -11

94 lines not on this page · raw message

arch/arm/boot/dts/imx6q-phytec-pfla02.dtsi +33 -11

92 lines not on this page · raw message

arch/arm/boot/dts/imx6q-sabrelite.dts +41 -9

110 lines not on this page · raw message

arch/arm/boot/dts/imx6q-sbc6x.dts +25 -4
index ee6addf..cc062ec 100644
@@ -17,28 +17,49 @@
 	};
 };
 
+
 &fec {
 	pinctrl-names = "default";
-	pinctrl-0 = <&pinctrl_enet_1>;
+	pinctrl-0 = <&pinctrl_enet>;
 	phy-mode = "rgmii";
 	status = "okay";
 };
 
+&iomuxc {
+	imx6q_sbc6x {
+		pinctrl_enet: enetgrp {
+			fsl,pins = <MX6QDL_ENET_PINGRP1>;
+		};
+
+		pinctrl_uart1: uart1grp {
+			fsl,pins = <MX6QDL_UART1_PINGRP1>;
+		};
+
+		pinctrl_usbotg: usbotggrp {
+			fsl,pins = <MX6QDL_USBOTG_PINGRP1>;
+		};
+
+		pinctrl_usdhc3: usdhc3grp {
+			fsl,pins = <MX6QDL_USDHC3_PINGRP2>;
+		};
+	};
+};
+
 &uart1 {
 	pinctrl-names = "default";
-	pinctrl-0 = <&pinctrl_uart1_1>;
+	pinctrl-0 = <&pinctrl_uart1>;
 	status = "okay";
 };
 
 &usbotg {
 	pinctrl-names = "default";
-	pinctrl-0 = <&pinctrl_usbotg_1>;
+	pinctrl-0 = <&pinctrl_usbotg>;
 	disable-over-current;
 	status = "okay";
 };
 
 &usdhc3 {
 	pinctrl-names = "default";
-	pinctrl-0 = <&pinctrl_usdhc3_2>;
+	pinctrl-0 = <&pinctrl_usdhc3>;
 	status = "okay";
 };
arch/arm/boot/dts/imx6q-udoo.dts +14 -2
index 6e1ccdc..04a36ef 100644
@@ -21,19 +21,31 @@
 	};
 };
 
+&iomuxc {
+	imx6q_udoo {
+		pinctrl_uart2: uart2grp {
+			fsl,pins = <MX6QDL_UART2_PINGRP1>;
+		};
+
+		pinctrl_usdhc3: usdhc3grp {
+			fsl,pins = <MX6QDL_USDHC3_PINGRP2>;
+		};
+	};
+};
+
 &sata {
 	status = "okay";
 };
 
 &uart2 {
 	pinctrl-names = "default";
-	pinctrl-0 = <&pinctrl_uart2_1>;
+	pinctrl-0 = <&pinctrl_uart2>;
 	status = "okay";
 };
 
 &usdhc3 {
 	pinctrl-names = "default";
-	pinctrl-0 = <&pinctrl_usdhc3_2>;
+	pinctrl-0 = <&pinctrl_usdhc3>;
 	non-removable;
 	status = "okay";
 };
arch/arm/boot/dts/imx6q.dtsi +1 -0
index f024ef2..5b92750 100644
@@ -9,6 +9,7 @@
  */
 
 #include "imx6q-pinfunc.h"
+#include "imx6qdl-pingrp.h"
 #include "imx6qdl.dtsi"
 
 / {
arch/arm/boot/dts/imx6qdl-pingrp.h new file +503 -0

503 lines not on this page · raw message

arch/arm/boot/dts/imx6qdl-sabreauto.dtsi +46 -12

109 lines not on this page · raw message

arch/arm/boot/dts/imx6qdl-sabresd.dtsi +51 -11

134 lines not on this page · raw message

arch/arm/boot/dts/imx6qdl-wandboard.dtsi +51 -11

132 lines not on this page · raw message

arch/arm/boot/dts/imx6qdl.dtsi +0 -738

744 lines not on this page · raw message

Signature · 1 line
1.7.9.5
#1 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Hi Shawn,

Quoted 18 lines · Shawn Guo

Currently, all pinctrl setting nodes are defined in <soc>.dtsi, so that boards that share the same pinctrl setting do not have to define it time and time again in <board>.dts. However, along with the devices and use cases being added continuously, the pinctrl setting nodes under iomuxc becomes more than expected. This bloats device tree blob for particular board unnecessarily since only a small subset of those pinctrl setting nodes will be used by the board. It impacts not only the DTB file size but also the run-time device tree lookup efficiency.

The patch proposes a solution to avoid this device tree bloating problem while still keeping boards share the common pinctrl setting data by using DTC macro support. It creates imx6qdl-pingrp.h and move all those pinctrl setting data into there as macro definitions. The <board>.dts will instead define the pinctrl setting nodes that are necessary for the board by referring to the macros in imx6qdl-pingrp.h, so that only the pinctrl setting data that will be used by the board will get compiled into the DTB for the board.

Quoted 7 lines

With the changes, the pinctrl setting nodes becomes local to particular board, and it makes no sense to continue numbering the setting for given peripheral. Thus, all the pinctrl phandler name gets updated to have only peripheral name in there.

Using imx6q-sabresd.dtb today as example, the change shrinks the file from 37KiB to 26KiB.

We have a similar problem, some boards need no pin mux at all (it's already done by the boot loader), some boards need some pin mux. Doing it the way it's done at the moment, we have ~10k dead pin mux stuff in the dtb for the boards without any pin mux. As you describe above.

While thinking about several options to solve this, we found

http://permalink.gmane.org/gmane.linux.drivers.devicetree/19694

Adding then something like

&iomuxc {
        status = "disabled";
};

to the dts which don't want any pin mux and then using the --strip-disabled option dramatically reduces the resulting dtb size.

There are several pros and cons to consider, though:

Pros:

* You remove additional unused parts, too. Not only the pin mux. I.e. the resulting dtb should be even smaller as with only removing the pin mux.

Cons:

* Non mainline dtc patch
* Not boot tested, yet. I have tried this only regarding the size of the resulting dtb. I haven't tested if the resulting dtb still boots, i.e. if everything essential isn't removed.

Best regards

Dirk

author
#2 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 10 lines · Dirk Behme

We have a similar problem, some boards need no pin mux at all (it's already done by the boot loader), some boards need some pin mux. Doing it the way it's done at the moment, we have ~10k dead pin mux stuff in the dtb for the boards without any pin mux. As you describe above.

While thinking about several options to solve this, we found

http://permalink.gmane.org/gmane.linux.drivers.devicetree/19694

Quoted 8 lines

Adding then something like

&iomuxc {
       status = "disabled";
};

to the dts which don't want any pin mux and then using the --strip-disabled option dramatically reduces the resulting dtb size.

Ah, I like the approach. But it does not actually address my problem, which is most of boards need iomuxc node but only a small set of pinctrl sub-nodes under it.

It does help to solve another concern from Matt regarding to DTB size, that is those nodes with 'disabled' property do not have to be in DTB to consume space and hurt run-time lookup efficiency. This is the Pros point that you listed below.

Quoted 8 lines

There are several pros and cons to consider, though:

Pros:

* You remove additional unused parts, too. Not only the pin mux. I.e. the resulting dtb should be even smaller as with only removing the pin mux.

Quoted 6 lines

Cons:

* Non mainline dtc patch
* Not boot tested, yet. I have tried this only regarding the size of the resulting dtb. I haven't tested if the resulting dtb still boots, i.e. if everything essential isn't removed.

They do not sound like any real Cons points to me :)

Shawn

maintainer
#3 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

In reply to Shawn Guo
Quoted 18 lines · Shawn Guo

Currently, all pinctrl setting nodes are defined in <soc>.dtsi, so that boards that share the same pinctrl setting do not have to define it time and time again in <board>.dts. However, along with the devices and use cases being added continuously, the pinctrl setting nodes under iomuxc becomes more than expected. This bloats device tree blob for particular board unnecessarily since only a small subset of those pinctrl setting nodes will be used by the board. It impacts not only the DTB file size but also the run-time device tree lookup efficiency.

The patch proposes a solution to avoid this device tree bloating problem while still keeping boards share the common pinctrl setting data by using DTC macro support. It creates imx6qdl-pingrp.h and move all those pinctrl setting data into there as macro definitions. The <board>.dts will instead define the pinctrl setting nodes that are necessary for the board by referring to the macros in imx6qdl-pingrp.h, so that only the pinctrl setting data that will be used by the board will get compiled into the DTB for the board.

Quoted 7 lines

With the changes, the pinctrl setting nodes becomes local to particular board, and it makes no sense to continue numbering the setting for given peripheral. Thus, all the pinctrl phandler name gets updated to have only peripheral name in there.

Using imx6q-sabresd.dtb today as example, the change shrinks the file from 37KiB to 26KiB.

Nice. Having all these unused nodes in the i.MX DTs concerned me aswell.

Quoted 8 lines

---
Russell,

With this approach, I will have no concern with adding another full group of USDHC pins with only one different from the existing group.

Quoted 11 lines
#define MX6QDL_USDHC1_PINGRP1_DAT3CD \
        MX6QDL_PAD_SD1_CMD__SD1_CMD    0x17059 \
        MX6QDL_PAD_SD1_CLK__SD1_CLK    0x10059 \
        MX6QDL_PAD_SD1_DAT0__SD1_DATA0 0x17059 \
        MX6QDL_PAD_SD1_DAT1__SD1_DATA1 0x17059 \
        MX6QDL_PAD_SD1_DAT2__SD1_DATA2 0x17059 \
        MX6QDL_PAD_SD1_DAT3__SD1_DATA3 0x13059 \
        MX6QDL_PAD_NANDF_D0__SD1_DATA4 0x17059 \
        MX6QDL_PAD_NANDF_D1__SD1_DATA5 0x17059 \
        MX6QDL_PAD_NANDF_D2__SD1_DATA6 0x17059 \
        MX6QDL_PAD_NANDF_D3__SD1_DATA7 0x17059

What I'm thinking about now for some time is whether I like these defines. They add another level of indirection and still the pinctrl settings cannot be adjusted by board dts files.
On the other hand nobody forces us to actually use these defines. If we realize that these defines are too inconvenient we could easily put the inidividual pins into the board dts files and remove the defines later.

So:

Sascha

Signature · 4 lines
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
#4 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 3 lines · Sascha Hauer

So:

Ok, I'll wait for this to hit mainline before re-spinning the Carrier-1 DT files. That's going to upset those who want to see Carrier-1 support in mainline at the next merge window (like Olof) and out mainline support for this board into 2014, but I don't see any other alternative.

#5 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

I can follow this up with i.MX51/i.MX53 if nobody is already doing that..

Quoted 9 lines · Russell King - ARM Linux
Quoted 3 lines · Sascha Hauer

So:

Ok, I'll wait for this to hit mainline before re-spinning the Carrier-1 DT files. That's going to upset those who want to see Carrier-1 support in mainline at the next merge window (like Olof) and out mainline support for this board into 2014, but I don't see any other alternative.

#6 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 7 lines · Matt Sealey

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

The issue I have is that this pushes support for Carrier-1 through to _six_ months before it appears in a mainline -final kernel - the timeline ends up looking like this:

- Prepare and merge this patch to allow DAT3 to be configured in the merge window starting next week.
- Rebase Carrier-1 changes on that when it hits mainline
- Wait for 3.13-rc1
- Get Carrier-1 DT files merged into the IMX tree for the following merge window at some point
- Wait for v3.13 to be released about three months from now.
- Carrier-1 support gets merged into mainline
- Carrier-1 support appears in v3.14 three months after that.

That's an awfully long turn-around time given that the DT description is functional today, and has been for the last month, and the only thing that's really blocking it is the problems of that DAT3 signal.

#7 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 23 lines · Russell King - ARM Linux
Quoted 7 lines · Matt Sealey

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

The issue I have is that this pushes support for Carrier-1 through to _six_ months before it appears in a mainline -final kernel - the timeline ends up looking like this:

- Prepare and merge this patch to allow DAT3 to be configured in the merge window starting next week.
- Rebase Carrier-1 changes on that when it hits mainline
- Wait for 3.13-rc1
- Get Carrier-1 DT files merged into the IMX tree for the following merge window at some point
- Wait for v3.13 to be released about three months from now.
- Carrier-1 support gets merged into mainline
- Carrier-1 support appears in v3.14 three months after that.

Quoted 3 lines

That's an awfully long turn-around time given that the DT description is functional today, and has been for the last month, and the only thing that's really blocking it is the problems of that DAT3 signal.

The DT description isn't changing at all - what comes out is going to work either way. I don't understand Shawn's rejection of adding another pinctrl definition for your board, as what the kernel parses at the end of the day will be almost identical.

It's a weird issue, but there are two big things that bloat device trees right now.

~

Problem #1: since "common" pinctrl definitions got moved into the .dtsi files, every board including that .dtsi file gets every pinctrl definition for every other board, whether they use the component that it's for or not. Each .dts for each board would just refer to the phandle for the definition in the .dtsi file, which made the .dts files shorter.

That means for i.MX6 which has

* 4 SDHC
* 5 UART
* 5 eCSPI controllers

Most boards use at most a couple of those. For example on SabreSD:

* 3x SDHC (there are two SD card slots an an eMMC)
* 1x SPI
* 1x UART.

Common pinctrls are present for at all 4 SDHC, 4 of the UART (nobody uses one of them), and a couple SPI controllers. That means SabreSD contained pinctrl definitions for 1 extra SD card, 3 extra UARTs, and 2 or 3 SPI controllers. Less source code meant bigger binaries.

You can go see it today; make imx6q-sabresd.dtb and then "dtc -I dtb -O dts imx6q-sabresd.dtb | less" and search for pinctrl nodes. There are far, far too many for the SabreSD board.

So that's problem #1 which this patching is solving. Now, when you build imx6q-sabresd.dtb you will only have pinctrl for the buses and IO it actually uses, and in this case, that's a good few KiB of data that was just redundant.. now gone away.

~

Problem #2: every node with status = "disabled" is bundled into the device tree. This means for SabreSD even though now only the relevant pinctrl nodes are present per-board, it still has nodes for 1 extra SD card, 4 extra UARTs and 4 SPI controllers that aren't even muxed out, and Linux ignores them anyway. This soaks up another goodly few KiB.

~

In total for Problem #1 and Problem #2 still existing, this is the difference between a 46KiB device tree and a 25KiB device tree blob. What HASN'T changed is how this all compiles down; it's in the .dts file and not the .dtsi file, and the dts file uses some macros to make life easier. This leads to slightly more device tree source code, but a much smaller binary. So you can just add:

&usdhc3 {
pinctrl-names = "default";
pinctrl-0 = <&russellsdcard>;
};

&iomuxc {

   usdhc3 {
      russellsdcard: russellgrp {
         fsl,pins = <MX6QDL_PAD1_FUNC 0x82828
                          MX6QDL_PAD2_FUNC 0x82828>;
      };

};
};

What you'll be able to do later is:

&iomuxc {
usdhc3 {

   russellsdcard: russellgrp {
        fsl,pins = <MX6QDL_WHATEVER_DEV_FUNC_VARIANT>;

};
};
};

And if Shawn doesn't like it, Shawn can go fish, because you're both just being stubborn for different reasons :)

The important thing to note is that the label (russellsdcard) is totally internal to the device tree compiler, it gets resolved to be a phandle eventually pointing to the node containing fsl,pins. The container node name (russellgrp) is also totally irrelevant here, and the container node above it (usdhc3) is completely redundant except to make the source (both in 'source' form and in 'disassembled' form) look pretty.

As long as your binary comes out the same and it parses the same things out, there's no waiting necessary because there's no ABI change at all. It may cause a tiny bit of churn later to basically reduce the amount of code (moving it from your file to pingrp.h) - is that what you're waiting on?

I'd say screw it and just get the board support in there. Code cleanups are a fact of life. This'd be like waiting 2 kernels for a whitespace change.

Thanks,
Matt Sealey <[email protected]>

author
#8 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

In reply to Matt Sealey
Quoted 9 lines · Matt Sealey

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

I can follow this up with i.MX51/i.MX53 if nobody is already doing that..

I'm taking care of them as a move of while IMX family.

Shawn

#9 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 9 lines · Shawn Guo
Quoted 8 lines · Matt Sealey

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

Quoted 3 lines
Quoted 1 line

I can follow this up with i.MX51/i.MX53 if nobody is already doing that..

I'm taking care of them as a move of while IMX family.

I'd quite like to push the Efika MX trees at the state they're at, sometime in the near future; if this work is going to take a while could I just do it and submit it and see your objections?

I have no merge deadline like Russell, but I would like to put some closure on some parts of my local tree and sync upstream. My Trello cards are getting unmanageable ;)

Thanks,
Matt Sealey <[email protected]>

author
#10 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 10 lines · Matt Sealey
Quoted 9 lines · Shawn Guo
Quoted 8 lines · Matt Sealey

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

I can follow this up with i.MX51/i.MX53 if nobody is already doing that..

I'm taking care of them as a move of while IMX family.

I'd quite like to push the Efika MX trees at the state they're at, sometime in the near future; if this work is going to take a while could I just do it and submit it and see your objections?

http://thread.gmane.org/gmane.linux.ports.arm.kernel/277649

It's on my for-next branch for testing now.

Shawn

#11 lore raw

RFC PATCH 1/1 ARM: dts: imx6qdl: make pinctrl nodes board specific

Quoted 15 lines · Shawn Guo
Quoted 14 lines · Matt Sealey
Quoted 12 lines · Shawn Guo
Quoted 9 lines · Matt Sealey

I suggested this to Shawn a couple weeks ago and got a bit of pushback.

There's probably a ton more to do in this regard to keep the DT per-board more concise, but this drops tree size by a respectable amount on i.MX51 too (just on a tree I have internally, it's something like 8.5KiB) and I can't see any redundant data in the ones he patched.

I can follow this up with i.MX51/i.MX53 if nobody is already doing that..

I'm taking care of them as a move of while IMX family.

Quoted 7 lines
Quoted 3 lines

I'd quite like to push the Efika MX trees at the state they're at, sometime in the near future; if this work is going to take a while could I just do it and submit it and see your objections?

http://thread.gmane.org/gmane.linux.ports.arm.kernel/277649

It's on my for-next branch for testing now.

Got it. Will jiggle everything to suit and just base against that.

Thanks,
Matt Sealey <[email protected]>