HDMI was flicker on demo kit when running stress on gpu

bill_tu · May 15, 2020, 10:00am

Wayne.

This week , we had updated patch on R32.3.1 and running . Ths issue was happen on 1~2 day. Can you help to check it.

Bill

WayneWWW · May 15, 2020, 10:47am

Hi Bill,

Could you use rel-32.4.2 and use that patch and see if you could still hit it?

Also, please do check the dmesg and make sure it is same error or new issue.

HuiW · May 19, 2020, 2:04am

Hi WayneWWW,

We are testing R32.4.2 with the patch.

However, the nvgpu.ko from we built is about 88M.
The original nvgpu.ko is around 2.4M.

We did get the same size difference on R32.4.2 and R32.3.1.

Does you have the same size difference?

Thank you,

WayneWWW · May 19, 2020, 2:10am

Hi HuiW,

That big size difference is because you didn’t strip the symobol of your nvgpu.ko.

https://docs.nvidia.com/jetson/l4t/index.html#page/Tegra%2520Linux%2520Driver%2520Package%2520Development%2520Guide%2Fkernel_custom.html%23wwpID0EUHA

HuiW · May 19, 2020, 2:24am

Hi WayneWWW,

Thank you for your prompt support.
So the ko file with symbol should not affect the patch test, right?

Thanks,

WayneWWW · May 19, 2020, 2:26am

It should not. But better stripping it.

bill_tu · May 20, 2020, 5:27am

dear Wayne.

 We had try to use R32.4.2(update patch) and same issue was happen.

Bill

WayneWWW · May 20, 2020, 5:32am

Hi Bill,

As I said, please also share the dmesg so that we can prove this issue is same as before.
Also, how much is the repro rate? How long it took to hit this issue?

bill_tu · May 22, 2020, 1:35am

error.log (109.7 KB)

Dear Wayne.

For R34.4.2, we  was duplication ths issue on the one day.

   Bill .

WayneWWW · May 22, 2020, 2:32am

Hi Bill,

Thanks for sharing. Also want to make sure you’ve replaced the nvgpu.ko, right?

bill_tu · May 22, 2020, 3:00am

dear Wayne.

yes, we use new nvgpu.ko to test and test reult.

Bill

WayneWWW · May 22, 2020, 3:36am

We will check. Thanks for report.

WayneWWW · May 26, 2020, 4:12am

Hi Bill,

We verified this patch on rel-32.4.2 again. Ran for 22 hours and no error is shown.
I notice your device is in very high temperature. Would you might put a fan there and try again? I don’t want any throttling affects the test.

Also, could you add a print in the nvgpu so that we can confirm that you really replace the kernel module.

WayneWWW · June 1, 2020, 9:02am

Hi Bill and HuiW,

Have you checked the temperature issue?

bill_tu · June 4, 2020, 3:15am

dear Wayne.

We had put a fan and reduce temperature . but the issue was haapen on 2 day .
Billerror.log (128.3 KB)

WayneWWW · June 4, 2020, 3:17am

Hi Bill,

Does this always need to take 2 days to reproduce? So far we have not tried such long time before.

bill_tu · June 4, 2020, 6:23am

dear Wayne.

We had try again on 5 hours and duplicate it (put on Fan) .

Bill

WayneWWW · June 4, 2020, 8:04am

Hi Bill,

Could you try this issue and patch on multiple Nano device? For example, B01/A02 and production module. It is fine with custom carrier board for this test.

I just checked and notice our test is also running for 2 days but no issue.

bill_tu · June 10, 2020, 5:51am

dear Wayne.

we had try two B01 module and duplication same issue . We use run 4H (power mode=MaxN) and can be duplication it.

Bill

WayneWWW · June 10, 2020, 6:29am

Do you have A02 module to do the test?