[PATCH v5 4/4] arm64: dts: qcom: add HDMI nodes for msm8998

[Marc Gonzalez]

On 23/07/2024 15:43, Konrad Dybcio wrote:

> On 23.07.2024 3:38 PM, Marc Gonzalez wrote:
>
>> On 23/07/2024 15:08, Konrad Dybcio wrote:
>>
>>> On 23.07.2024 2:57 PM, Marc Gonzalez wrote:
>>>
>>>> On 23/07/2024 13:45, Konrad Dybcio wrote:
>>>>
>>>>> On 23.07.2024 11:59 AM, Dmitry Baryshkov wrote:
>>>>>
>>>>>> On Tue, 23 Jul 2024 at 12:48, Marc Gonzalez wrote:
>>>>>>
>>>>>>> On 16/07/2024 18:37, Dmitry Baryshkov wrote:
>>>>>>>
>>>>>>>> No, that's fine. It is the SMMU issue that Konrad has been asking you
>>>>>>>> to take a look at.
>>>>>>>
>>>>>>> Context:
>>>>>>>
>>>>>>> [    4.911422] arm-smmu cd00000.iommu: FSR    = 00000402 [Format=2 TF], SID=0x0
>>>>>>> [    4.923353] arm-smmu cd00000.iommu: FSYNR0 = 00000021 [S1CBNDX=0 PNU PLVL=1]
>>>>>>> [    4.927893] arm-smmu cd00000.iommu: FSR    = 00000402 [Format=2 TF], SID=0x0
>>>>>>> [    4.941928] arm-smmu cd00000.iommu: FSYNR0 = 00000021 [S1CBNDX=0 PNU PLVL=1]
>>>>>>> [    4.944438] arm-smmu cd00000.iommu: FSR    = 00000402 [Format=2 TF], SID=0x0
>>>>>>> [    4.956013] arm-smmu cd00000.iommu: FSYNR0 = 00000021 [S1CBNDX=0 PNU PLVL=1]
>>>>>>> [    4.961055] arm-smmu cd00000.iommu: FSR    = 00000402 [Format=2 TF], SID=0x0
>>>>>>> [    4.974565] arm-smmu cd00000.iommu: FSYNR0 = 00000021 [S1CBNDX=0 PNU PLVL=1]
>>>>>>> [    4.977628] arm-smmu cd00000.iommu: FSR    = 00000402 [Format=2 TF], SID=0x0
>>>>>>> [    4.989670] arm-smmu cd00000.iommu: FSYNR0 = 00000021 [S1CBNDX=0 PNU PLVL=1]
>>>>>>>
>>>>>>>
>>>>>>> As I mentioned, I don't think I've ever seen issues from cd00000.iommu
>>>>>>> on my board.
>>>>>>
>>>>>> Interestingly enough, I can also see iommu errors during WiFi startup
>>>>>> / shutdown on msm8998 / miix630. This leads me to thinking that it
>>>>>> well might be that there is a missing quirk in the iommu driver.
>>>>>>
>>>>>>> I can test a reboot loop for a few hours, to see if anything shows up.
>>>>>>
>>>>>> Yes, please.
>>>>>
>>>>> Yeah I do trust you Marc that it actually works for you and I'm not
>>>>> gonna delay this series because of that, but please go ahead and
>>>>> reboot-loop your board
>>>>>
>>>>> 8998/660 is """famous""" for it's iommu problems
>>>>
>>>> [   20.501062] arm-smmu 16c0000.iommu: Unhandled context fault: fsr=0x402, iova=0x00000000, fsynr=0x1, cbfrsynra=0x1900, cb=0
>>>>
>>>> I get the above warning pretty reliably.
>>>> I don't think it's related to the issue(s) you mentioned.
>>>> System just keeps plodding along.
>>>
>>> Yeah that one's "fine"
>>
>> I booted 40 times in a loop.
>>
>> `grep -a -i FSYNR console.logs` just returns the same 16c0000.iommu
>> "Unhandled context fault" message 76 times (as above).
>>
>> NB: I have maxcpus=1 set in bootargs.
>>
>> Could the iommu issue be a race condition, NOT triggered when code
>> runs with less parallelism?
> 
> No clue, can you try without maxcpus=1?

Same behavior without maxcpus=1

40 boots, no panics, no FSYNR other than 16c0000.iommu

> The thing will likely run slower (because reasons), but shouldn't
> explode

That makes sense!

- Hey, boot is slow. What can we do to make it slower?
- Well, just add a bunch of cores running in parallel, that will get the job done!

As a matter of fact, trying to boot to command-line with
maxcpus=1 causes the system to lock up & reboot.
I had to add a systemd script to enable some cores at init.
Some qcom daemon must be locking a core & expect progress
from another process.

Regards