Message-ID: <28c72c1a-df7d-db05-f32f-23b170b96db5@linaro.org>
Date:   Sat, 25 Feb 2023 11:00:48 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.8.0
Subject: Re: [PATCH v1 1/2] dt-bindings: PCI: dwc: Add snps,skip-wait-link-up
To:     Sajid Dalvi <sdalvi@google.com>, Jingoo Han <jingoohan1@gmail.com>,
        Gustavo Pimentel <gustavo.pimentel@synopsys.com>,
        Lorenzo Pieralisi <lpieralisi@kernel.org>,
        =?UTF-8?Q?Krzysztof_Wilczy=c5=84ski?= <kw@linux.com>,
        Rob Herring <robh@kernel.org>,
        Bjorn Helgaas <bhelgaas@google.com>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
Cc:     kernel-team@android.com, linux-pci@vger.kernel.org,
        devicetree@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20230224195749.818282-1-sdalvi@google.com>
 <20230224195749.818282-2-sdalvi@google.com>
 <b50ab99f-e307-3a66-9198-85a71b012e5e@linaro.org>
 <CAEbtx1k-7TJPcd3+cueRoKLJcoUQLfF6nfOQFVfzB0YCUrbtqg@mail.gmail.com>
 <2e4964c1-0831-c156-3372-81a56f8d623e@linaro.org>
 <CAEbtx1=EGC+4LfSz+HqGzjJvF2O1vsB7a42ESTqbU8p5N-yUHg@mail.gmail.com>
Content-Language: en-US
From:   Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
In-Reply-To: <CAEbtx1=EGC+4LfSz+HqGzjJvF2O1vsB7a42ESTqbU8p5N-yUHg@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 24/02/2023 23:09, Sajid Dalvi wrote:
> On Fri, Feb 24, 2023 at 3:29 PM Krzysztof Kozlowski
> <krzysztof.kozlowski@linaro.org> wrote:
>>
>> On 24/02/2023 22:27, Sajid Dalvi wrote:
>>> On Fri, Feb 24, 2023 at 2:40 PM Krzysztof Kozlowski
>>> <krzysztof.kozlowski@linaro.org> wrote:
>>>>
>>>> On 24/02/2023 20:57, Sajid Dalvi wrote:
>>>>> When the Root Complex is probed, the default behavior is to spin in a loop
>>>>> waiting for the link to come up. In some systems the link is not brought up
>>>>> during probe, but later in the context of an end-point turning on.
>>>>> This property will allow the loop to be skipped.
>>>>>
>>>>> Signed-off-by: Sajid Dalvi <sdalvi@google.com>
>>>>> ---
>>>>
>>>> Thank you for your patch. There is something to discuss/improve.
>>>>
>>>>>  Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml | 8 ++++++++
>>>>>  1 file changed, 8 insertions(+)
>>>>>
>>>>> diff --git a/Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml b/Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml
>>>>> index 1a83f0f65f19..0b8950a73b7e 100644
>>>>> --- a/Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml
>>>>> +++ b/Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml
>>>>> @@ -197,6 +197,14 @@ properties:
>>>>>        - contains:
>>>>>            const: msi
>>>>>
>>>>> +  snps,skip-wait-link-up:
>>>>> +    $ref: /schemas/types.yaml#/definitions/flag
>>>>> +    description:
>>>>> +      When the Root Complex is probed, the default behavior is to spin in a
>>>>> +      loop waiting for the link to come up. In some systems the link is not
>>>>> +      brought up during probe, but later in the context of an end-point turning
>>>>> +      on. This property will allow the loop to be skipped.
>>>>
>>>> I fail to see how probe behavior is related to properties of hardware.
>>>> You describe OS behavior, not hardware. This does not look like
>>>> belonging to DT.
>>>>
>>>>
>>>> Best regards,
>>>> Krzysztof
>>>
>>> Thanks for your response Krzysztof.
>>> The hardware configuration of the system determines whether an
>>> endpoint device is available during host init. If it isn't available
>>> on a particular and dedicated pcie interface, we should skip waiting
>>> for the link to be up. For other interfaces, possibly even on the same
>>> system, where a device is present or maybe present we should wait for
>>> the link to come up.
>>
>> Keep discussions public.
>>
>> Your commit and property description mentions probe, which is nothing
>> related to hardware. Why the device would not be available during host
>> init (I understand we do not talk about hotplug as it is already
>> supported by Linux) in a way it is hardware property, not OS?
>>
>> Best regards,
>> Krzysztof
>>
> 
> + everyone else I mistakenly didn't reply to earlier
> 
> If I understand you correctly, the usage of probe is misleading
> because it doesn't have anything to do with the hardware.
> So your recommendation is to replace probe with device init, in the
> description of the property and the commit message?

No, I asked there a question for which we need answer.

device init is also OS task... You need to explain why this is a
property of hardware, not OS behavior.

Best regards,
Krzysztof