Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <3d9d545d-a620-85f6-b7bd-d57a8729f818@linaro.org>
Date:   Fri, 5 May 2023 14:23:20 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.10.1
Subject: Re: [PATCH 7/7] dt-bindings: dmaengine: pl330: Add new quirks
Content-Language: en-US
To:     Joy Chakraborty <joychakr@google.com>
Cc:     Vinod Koul <vkoul@kernel.org>, Rob Herring <robh+dt@kernel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
        dmaengine@vger.kernel.org, linux-kernel@vger.kernel.org,
        devicetree@vger.kernel.org, manugautam@google.com,
        danielmentz@google.com, sjadavani@google.com
References: <20230504145737.286444-1-joychakr@google.com>
 <20230504145737.286444-8-joychakr@google.com>
 <78616bc1-8d9e-4a1c-70d6-ad62c2cfa8a8@linaro.org>
 <CAOSNQF15UN2Rckes55UHxbUvN1PJcbj9aWirVGSLDOs5Y5EPnQ@mail.gmail.com>
From:   Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
In-Reply-To: <CAOSNQF15UN2Rckes55UHxbUvN1PJcbj9aWirVGSLDOs5Y5EPnQ@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 05/05/2023 11:44, Joy Chakraborty wrote:
> On Thu, May 4, 2023 at 8:38 PM Krzysztof Kozlowski
> <krzysztof.kozlowski@linaro.org> wrote:
>>
>> On 04/05/2023 16:57, Joy Chakraborty wrote:
>>> Add 2 new quirks added to the driver "arm,pl330-optimize-dev2mem-axsize"
>>> and "arm,pl330-periph-single-dregs"
>>
>> This we can see from the diff. You need to answer why?
>>
> 
> Sure will change it to:
> "
> Addition of following quirks :
> - "arm,pl330-periph-use-diff-axsize"
>    AxSize of transactions to peripherals are limited by the peripheral
> address width which inturn limits the AxSize used for transactions
> towards memory.
>    This quirk will make transactions to memory use the maximum
> possible bus width(AxSize), store data in MFIFO and use narrow
> multi-beat transactions to move data to peripherals.
>    This only applies to transfers between memory and peripherals where
> bus widths available are different for memory and the peripheral.
> - "arm,pl330-periph-complete-with-singles" :
>    When transfer sizes are not a multiple of a block of burst
> transfers (AxLen * AxSize configured at the peripheral), certain
> peripherals might choose not to set the burst request at the
> peripheral request interface of the DMA.
>    This quirk moves the remaining bytes to the peripheral using single
> transactions.
> "

This does not answer why. You just copied again the patch contents.

> 
>>>
>>> Signed-off-by: Joy Chakraborty <joychakr@google.com>
>>> ---
>>>  Documentation/devicetree/bindings/dma/arm,pl330.yaml | 8 ++++++++
>>>  1 file changed, 8 insertions(+)
>>>
>>> diff --git a/Documentation/devicetree/bindings/dma/arm,pl330.yaml b/Documentation/devicetree/bindings/dma/arm,pl330.yaml
>>> index 4a3dd6f5309b..0499a7fba88d 100644
>>> --- a/Documentation/devicetree/bindings/dma/arm,pl330.yaml
>>> +++ b/Documentation/devicetree/bindings/dma/arm,pl330.yaml
>>> @@ -53,6 +53,14 @@ properties:
>>>      type: boolean
>>>      description: quirk for performing burst transfer only
>>>
>>> +  arm,pl330-optimize-dev2mem-axsize:
>>> +    type: boolean
>>> +    description: quirk for optimizing AxSize used between dev<->mem
>>
>> This tells me nothing... Neither what it is about nor why this is
>> property of a board or PL330 hardware implementation. Please describe
>> hardware, not drivers.
>>
> 
> Will change the name to "arm,pl330-periph-use-diff-axsize" and add description:
> "
> Quirk to use different AxSize for bursts while accessing source and
> destination when moving data between memory and peripheral.
> Maximum possible bus width is used as AxSize for transactions towards
> memory and transactions towards peripherals use AxSize as per
> peripheral address width.
> "

Still no answer. Why this is property of a board or PL330 hardware
implementation?

I also asked to describe hardware but I still see "quirk to ...". We use
"quirk" as concept in Linux driver. Describe the hardware, not Linux driver.


> 
>>> +
>>> +  arm,pl330-periph-single-dregs:
>>> +    type: boolean
>>> +    description: quirk for using dma-singles for peripherals in _dregs()
>>
>> Same concerns.
>>
> 
> Will change the name to  "arm,pl330-periph-complete-with-singles" and
> add description:
> "
> Quirk to use dma singles n times instead of an n beat burst to
> complete a transfer when the transfer size is not a multiple of the

No, how you wrote it sounds like driver. Don't add driver quirks to DT.

Best regards,
Krzysztof