Re: [PATCH v3 2/9] ARM: dts: nspire: Use syscon-reboot to handle restart

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 10/27/22 4:27 PM, Krzysztof Kozlowski wrote:
On 27/10/2022 17:07, Andrew Davis wrote:
On 10/27/22 2:33 PM, Krzysztof Kozlowski wrote:
On 27/10/2022 14:13, Andrew Davis wrote:
Writing this bit can be handled by the syscon-reboot driver.
Add this node to DT.

Signed-off-by: Andrew Davis <afd@xxxxxx>
Reviewed-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
Tested-by: Fabian Vogt <fabian@xxxxxxxxxxxxxx>
Reviewed-by: Fabian Vogt <fabian@xxxxxxxxxxxxxx>
---
   arch/arm/boot/dts/nspire.dtsi | 7 +++++++
   1 file changed, 7 insertions(+)

diff --git a/arch/arm/boot/dts/nspire.dtsi b/arch/arm/boot/dts/nspire.dtsi
index bb240e6a3a6f..48fbc9d533c3 100644
--- a/arch/arm/boot/dts/nspire.dtsi
+++ b/arch/arm/boot/dts/nspire.dtsi
@@ -172,7 +172,14 @@ rtc: rtc@90090000 {
   			};
misc: misc@900a0000 {
+				compatible = "ti,nspire-misc", "syscon", "simple-mfd";

You have syscon and simple-mfd, but bindings in patch #1 say only syscon.


I'm not following, are you just saying my wording in the patch message just
wasn't complete?

Your binding patch adds nspire compatible to the list of two items, so
you have two items in total - nspire followed by syscon.

What you implemented here is different.


Is there a list of three items I can add this compatible? If instead you
mean I should go make a new binding, just say so :)


Or are you saying something more about nodes that are both syscon and simple-mfd?
In that case, having both syscon and simple-mfd seems rather common, looks like
you added the rule for it[0].

Thinking on this, they almost represent the same thing. simple-mfd says "my child
nodes should be considered devices", why do we need that? Couldn't we simply state
that "syscon" node's children are always devices, I mean what else could they be,
syscon is an MFD after all (and lives in drivers/mfd/).

No, syscon is not an MFD. Syscon means system controller and alone it
does not have children.


The binding lives in devicetree/bindings/*mfd*/, it is mentioned as one
in devicetree/bindings/mfd/mfd.txt. If it is not an MFD then the bindings
are giving out mixed signals here..


"syscon" often just says, others can use the registers within this node, so as a
different option, make "syscon" a property of "simple-mfd" nodes. I'm seeing all
these examples of devices that should have been children of the "syscon" device,
but instead use

regmap = <&x>;
syscon = <&x>;

or similar and put the device node out somewhere random. And in those cases,
wouldn't it have been more correct to use the normal "reg" and "regions" to
define the registers belonging to the child node/device?..

Sorry, I do not follow. How this is even related to your patch?

Your bindings say A, DTS say B. A != B. This needs fixing.


I said it was compatible with "syscon", not that it is incompatible
with "simple-mfd" devices.

What I've done here gives no dtbs_check warnings and
"devicetree/bindings/mfd/mfd.txt" explicitly allows what I am doing.
Unless we do not consider the old bindings valid? If so, would you
like me to convert mfd.txt to yaml, just let me know.

Unless you are asking me what your device is in general. This I don't
really know, but if you want to use it as regmap provider for system
registers and as a parent of syscon-based reboot device, then your
device is syscon and simple-mfd. With a specific compatible. Was this
your question?


Yes, I would like to use it as a regmap provider, my question here is
a much more general one: why do I need to specify that in device tree?
That is not a hardware description, my hardware is not "regmap" hardware.
This "syscon" stuff feels like a bodge to make the Linux drivers and bus
frameworks interact the way we want.

I know at this point this has little to do with this series, but I'd like
to just think this out for a moment. The latest Devicetree Specification
talks about "simple-bus" as a special compatible that communicates that
child nodes with compatible strings need probed also. ("simple-mfd" seems
to be used the same way but without needing a "ranges" property..)

Both of these are properties of a node, not something a device is "compatible"
with. "compatibles" are also supposed to be listed "from most specific to
most general", so which is more specific, "simple-mfd" or "syscon", etc..

Seems like Rob might agree[0], these are not really compatibles. We cant fix
history, but for new nodes, instead of growing the problem and forcing these to
be overloaded compatibles, we allow these to become new standard node properties.

For instance:

main_conf: syscon@43000000 {
	compatible = "ti,j721e-system-controller";
	reg = <0x0 0x43000000 0x0 0x20000>;

	simple-bus;
	syscon;

	...
};

Thoughts?

Thanks,
Andrew

[0] https://lore.kernel.org/all/CAL_JsqKiUcO76bo1GoepWM1TusJWoty_BRy2hFSgtEVMqtrvvQ@xxxxxxxxxxxxxx/

Best regards,
Krzysztof




[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux