Re: [PATCH v9 6/7] arm64: dts: qcom: ipq5332: Add PCIe related nodes
From: Krzysztof Kozlowski
Date: Thu Feb 06 2025 - 02:37:55 EST
On 06/02/2025 07:19, Varadarajan Narayanan wrote:
> On Wed, Feb 05, 2025 at 04:54:38PM +0100, Krzysztof Kozlowski wrote:
>> On 05/02/2025 16:53, Krzysztof Kozlowski wrote:
>>> On 05/02/2025 16:35, Varadarajan Narayanan wrote:
>>>> On Wed, Feb 05, 2025 at 02:47:13PM +0100, Krzysztof Kozlowski wrote:
>>>>> On 05/02/2025 11:49, Varadarajan Narayanan wrote:
>>>>>> On Mon, Feb 03, 2025 at 05:30:32PM +0100, Krzysztof Kozlowski wrote:
>>>>>>> On 28/01/2025 07:27, Varadarajan Narayanan wrote:
>>>>>>>>
>>>>>>>> @@ -479,6 +519,230 @@ frame@b128000 {
>>>>>>>> status = "disabled";
>>>>>>>> };
>>>>>>>> };
>>>>>>>> +
>>>>>>>> + pcie1: pcie@18000000 {
>>>>>>>> + compatible = "qcom,pcie-ipq5332", "qcom,pcie-ipq9574";
>>>>>>>> + reg = <0x00088000 0x3000>,
>>>>>>>
>>>>>>> So as Konrad pointed out now, this was never tested. It's not we who
>>>>>>> should run tests for you. It's you.
>>>>>>
>>>>>> This was tested and it did not flag an error since it is having the order
>>>>>> specified in the bindings. qcom,pcie.yaml has 4 reg specifications. Two of
>>>>>
>>>>>
>>>>> Hm, then please paste results of dtbs_check W=1 testing. Here.
>>>>>
>>>>> I am 100% sure you have there warning and I don't understand your
>>>>> reluctance to run the tests even after pointing it out by two people.
>>>>
>>>> I ran the tests. Not sure which portions to paste. Have attached the full
>>>> output just in case you are interested in some other detail. Please take a
>>>> look.
>>>>
>>>> Thanks
>>>> Varada
>>>>
>>>> $ grep ipq.*dtb dtbs-check.log
>>>
>>> Where is the command you have used?
>
> export ARCH=arm64
> export W=1
> export DT_CHECKER_FLAGS='-v -m'
> export DT_SCHEMA_FILES=qcom
> export CHECK_DTBS=y
>
> make -j 16 dtbs_check &> dtbs-check.log
>
No flags for DTC as I asked.
>> Although that might not matter - you skipped several warnings with your
>> grep. So maybe you need to fix your process, not sure.
>
> export W=1 is the problem. Kernel Makefile differentiates between 'W' being
> set from environment and from command line with this check
So just don't export. Command is:
make W=1 -j8 dtbs_check
>
> ifeq ("$(origin W)", "command line")
> KBUILD_EXTRA_WARN := $(W)
> endif
>
> I assumed similar to DT_SCHEMA_FILES and DT_CHECKER_FLAGS, W will also be
> taken. I was not aware of this differentiation, and the 'export W=1' never
> came into effect. I re-ran the command as below and see the warnings
>
> $ make W=1 -j 16 dtbs_check &> dtbs-check2.log
>
> $ grep Warning dtbs-check2.log | grep ipq.*dt
> arch/arm64/boot/dts/qcom/ipq5332.dtsi:523.24-625.5: Warning (simple_bus_reg): /soc@0/pcie@18000000: simple-bus unit address format error, expected "88000"
> arch/arm64/boot/dts/qcom/ipq5332.dtsi:627.24-729.5: Warning (simple_bus_reg): /soc@0/pcie@20000000: simple-bus unit address format error, expected "80000"
Oh look! What surprise, who could expect that...
Best regards,
Krzysztof