Message-ID: <af81f1e4-250b-4785-8b44-fc3e9753dba8@amlogic.com>
Date: Fri, 19 Apr 2024 09:51:05 +0800
From: Kelvin Zhang <kelvin.zhang@...ogic.com>
To: neil.armstrong@...aro.org,
 Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
 Philipp Zabel <p.zabel@...gutronix.de>, Rob Herring <robh@...nel.org>,
 Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
 Conor Dooley <conor+dt@...nel.org>, Kevin Hilman <khilman@...libre.com>,
 Jerome Brunet <jbrunet@...libre.com>,
 Martin Blumenstingl <martin.blumenstingl@...glemail.com>
Cc: devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
 linux-amlogic@...ts.infradead.org, linux-kernel@...r.kernel.org,
 Zelong Dong <zelong.dong@...ogic.com>
Subject: Re: [PATCH 1/3] dt-bindings: reset: Add Amlogic T7 Reset Controller



On 2024/4/19 05:03, neil.armstrong@...aro.org wrote:
> [ EXTERNAL EMAIL ]
> 
> On 17/04/2024 21:08, Krzysztof Kozlowski wrote:
>> On 16/04/2024 01:30, neil.armstrong@...aro.org wrote:
>>> On 15/04/2024 12:31, Kelvin Zhang wrote:
>>>>
>>>> On 2024/4/13 02:03, Krzysztof Kozlowski wrote:
>>>>> [ EXTERNAL EMAIL ]
>>>>>
>>>>> On 12/04/2024 19:57, Krzysztof Kozlowski wrote:
>>>>>> On 12/04/2024 15:12, Neil Armstrong wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 29/03/2024 20:39, Krzysztof Kozlowski wrote:
>>>>>>>> On 29/03/2024 10:17, Kelvin Zhang via B4 Relay wrote:
>>>>>>>>> From: Zelong Dong <zelong.dong@...ogic.com>
>>>>>>>>>
>>>>>>>>> Add a new compatible and the related header file
>>>>>>>>> for Amlogic T7 Reset Controller.
>>>>>>>>>
>>>>>>>>> Signed-off-by: Zelong Dong <zelong.dong@...ogic.com>
>>>>>>>>> Signed-off-by: Kelvin Zhang <kelvin.zhang@...ogic.com>
>>>>>>>>> ---
>>>>>>>>>     .../bindings/reset/amlogic,meson-reset.yaml        |   1 +
>>>>>>>>>     include/dt-bindings/reset/amlogic,t7-reset.h       | 197 
>>>>>>>>> +++++++++++++++++++++
>>>>>>>>>     2 files changed, 198 insertions(+)
>>>>>>>>>
>>>>>>>>> diff --git 
>>>>>>>>> a/Documentation/devicetree/bindings/reset/amlogic,meson-reset.yaml b/Documentation/devicetree/bindings/reset/amlogic,meson-reset.yaml
>>>>>>>>> index f0c6c0df0ce3..fefe343e5afe 100644
>>>>>>>>> --- 
>>>>>>>>> a/Documentation/devicetree/bindings/reset/amlogic,meson-reset.yaml
>>>>>>>>> +++ 
>>>>>>>>> b/Documentation/devicetree/bindings/reset/amlogic,meson-reset.yaml
>>>>>>>>> @@ -19,6 +19,7 @@ properties:
>>>>>>>>>           - amlogic,meson-a1-reset # Reset Controller on A1 and 
>>>>>>>>> compatible SoCs
>>>>>>>>>           - amlogic,meson-s4-reset # Reset Controller on S4 and 
>>>>>>>>> compatible SoCs
>>>>>>>>>           - amlogic,c3-reset # Reset Controller on C3 and 
>>>>>>>>> compatible SoCs
>>>>>>>>> +      - amlogic,t7-reset # Reset Controller on T7 and 
>>>>>>>>> compatible SoCs
>>>>>>>>>
>>>>>>>>
>>>>>>>> If there is going to be any resend, please drop the comment. 
>>>>>>>> It's not
>>>>>>>> really helpful and makes it trickier to read.
>>>>>>>>
>>>>>>>>>       reg:
>>>>>>>>>         maxItems: 1
>>>>>>>>> diff --git a/include/dt-bindings/reset/amlogic,t7-reset.h 
>>>>>>>>> b/include/dt-bindings/reset/amlogic,t7-reset.h
>>>>>>>>> new file mode 100644
>>>>>>>>> index 000000000000..ca4a832eeeec
>>>>>>>>> --- /dev/null
>>>>>>>>> +++ b/include/dt-bindings/reset/amlogic,t7-reset.h
>>>>>>>>> @@ -0,0 +1,197 @@
>>>>>>>>> +/* SPDX-License-Identifier: (GPL-2.0-only OR MIT) */
>>>>>>>>> +/*
>>>>>>>>> + * Copyright (c) 2024 Amlogic, Inc. All rights reserved.
>>>>>>>>> + */
>>>>>>>>> +
>>>>>>>>> +#ifndef _DT_BINDINGS_AMLOGIC_T7_RESET_H
>>>>>>>>> +#define _DT_BINDINGS_AMLOGIC_T7_RESET_H
>>>>>>>>> +
>>>>>>>>> +/* RESET0 */
>>>>>>>>> +/*                                        0-3     */
>>>>>>>>
>>>>>>>> I assume this matches existing drivers which do not use IDs but 
>>>>>>>> map the
>>>>>>>> binding to hardware value? I remember we talked about changing 
>>>>>>>> it, so if
>>>>>>>> something happened about this and it could be changed: please 
>>>>>>>> change.
>>>>>>>
>>>>>>> I'm not aware of such discussion, and I don't really see the 
>>>>>>> issue...
>>>>>>> thoses are IDs, and yes they match the Hardware offsets, and ?
>>>>>>
>>>>>> Bindings are not for hardware offsets/values/addresses. It's just 
>>>>>> not a
>>>>>> binding.
>>>>>>
>>>>>> I quickly looked at your driver patch and it confirms: not a binding.
>>>>>> Binding constant is used by the driver and DTS consumer.
>>>>>>
>>>>>> I am really sure we had this talk in the past, but could be I think
>>>>>> about different platform. Since this is not a binding, I do not think
>>>>>> claiming there is any ABI here is reasonable. Feel free to store them
>>>>>> with other hardware values, like in DTS headers etc. We already 
>>>>>> moved to
>>>>>> DTS headers several such "non-binding" constants.
>>>>>
>>>>> Un-acked.
>>>>>
>>>>> I looked at my archives and we did talk about it and you were CCed:
>>>>>
>>>>> https://lore.kernel.org/linux-devicetree/c088e01c-0714-82be-8347-6140daf56640@linaro.org/
>>>>> simple-reset is an exception.
>>>>>
>>>>> So to recap:
>>>>> That's not a binding. Don't add some real values to binding headers
>>>>> because it is not a binding then.
>>>
>>> So what's exactly a binding then?
>>
>> Binding headers is interface needed (necessary) between implementation
>> (like Linux drivers) and DTS.
>>
>>> random linear numbers that means nothing can be a binding
>>> but registers numbers can't be ? why ? I still don't understand, why 
>>> this suddenly gets problematic ?
>>
>> There is no interface here. Drivers don't use them. It's not "suddenly"
>> problematic, I commented on this year or two years ago and we also
>> started moving such header-abusers out of bindings.
>>
>>>
>>>>>
>>>>> https://lore.kernel.org/linux-devicetree/CAK8P3a1APzs74YTcZ=m43G3zrmwJZKcYSTvV5eDDQX-37UY7Tw@mail.gmail.com/
>>>>> https://lore.kernel.org/linux-devicetree/CAK8P3a0fDJQvGLEtG0fxLkG08Fh9V7LEMPsx4AaS+2Ldo_xWxw@mail.gmail.com/
>>>>> https://lore.kernel.org/linux-devicetree/b60f5fd2-dc48-9375-da1c-ffcfe8292683@linaro.org/
>>>>> https://lore.kernel.org/linux-devicetree/418c5f0c-5279-41f5-3705-345ec9a97ea2@linaro.org/
>>>>> https://lore.kernel.org/all/201401111415.29395.arnd@arndb.de/
>>>>>
>>>> Got it. Will delete amlogic,t7-reset.h and use the hardware numbers
>>>> directly in the DT. >
>>>> Hi Neil,
>>>> As you know, Amlogic reset controller is divided into several 
>>>> groups: reset0, reset1, ..., resetN. I'd like to discuss the 
>>>> rationality of splitting the one device node of reset controller 
>>>> into device nodes according to the groups. Then we can use the bit 
>>>> number within the 'resets' property.
>>>> reset0: reset-controller@...0 {
>>>> ...
>>>> };
>>>>
>>>> reset1: reset-controller@...4 {
>>>> ...
>>>> };
>>>> ...
>>>>
>>>> What do you think?
>>>
>>> No since you'll basically add a node per register, you need to add a 
>>> node for the while reset HW function, another
>>> solution would be to split the phandle arguments in 2, the first 
>>> first would be the reset bank, and the second one
>>> the reset line for the bank.
>>>
>>> But still it's a regression in readability to drop the macros, until 
>>> gpios or pins the reset number doesn't mean anything per se.
>>
>>
>> What stops you from putting the header in the DTS? Just like others are
>> doing?
> 
> Ok so now I understand, Kelvin just keep the header but move it in 
> arch/arm64/boot/dts/amlogic along the DT patch and drop it from the 
> bindings patch.
> 
Will do.
Thanks!

> Thanks,
> Neil
> 
>>
>> Best regards,
>> Krzysztof
>>
> 

Powered by blists - more mailing lists