HDMI was flicker on demo kit when running stress on gpu

dear Wayne.

  We had try to run CPU load and same issue happen. 
  CPU loading command: 1.export TOTAL_LOOP_TIME=99999999999999999999
                                             2. phoronix-test-suite stress-run pts/stress-ng

Bill

Our team is still checking the code of phoronix. Will update later.

Hi WayneWWW,

Thank you for your great support.
Is there any update from your team?

Thanks

Hi HuiW,

Sorry, we have some experimental patch but not yet fixed this issue.

Hi WayneWWW,

Thank you for your update.
Please let us know, if the patch is ready.

Thanks,

Hi HuiW,

Please try with this patch.

1e3befa.diff.zip (1.1 KB)

Hi WayneWWW,

Could you help check which L4T version is the patch for?

As I checked the source on L4T R32.3.1 and L4t R32.4.2, there is no drivers/gpu/nvgpu/common path.
Only drivers/gpu/drm/nouveau/nvkm/subdev/bios/pmu.c.

Thank you for any advice,

Do you see this path in your source?

nvgpu/drivers/gpu/nvgpu/common

Hi WayneWWW,

Thank you for your prompt support.
Yes. Got the path.

So is this patch for L4T R32.4.2 not R32.2.1?
There is no nvgpu_cg_blcg_enable_no_wait(g) on L4t R32.2.1.

We are still testing the stress test on L4T R32.4.2.

Thank you,

This patch is for rel-32.4.2.

Hi @WayneWWW,

I am experiencing the same gpu issues as in the referenced posts.

Could you please provide information/steps on how and where to apply this patch as I have not done this before?

Thank you.

Please refer to l4t development guide -> Kernel customization.

dear Wayne.

Our release BSP was R32.3.1 .So can be you help to porivide patch on R32.3.1 for us.

Bill

Hi Bill,

For rel-32.3.1, you need to use this patch too.
e45e7b5.diff.zip (1.5 KB)
Put this patch first and then the patch in previous comment.

Actually, I hope you could move to rel-32.4.2 to debug this issue. All our patch would be based on latest release instead of specific release in past.

Wow, that is a lot of stuff to do which I am not familiar with (yet).

Is there not an easier fix of applying a patched kernel to verify faster that it actually solves the issue?

Regards

Unfortunately, this is already the most fast way to patched the kerenl.

run the arm64 image inside of qemu. setting up qemu is not trivial but once you do it, easy to test.

@bill_tu and @HuiW

Have you resolved this issue with that patch?

Hi WayneWWW,

I’m afraid that the patch for L4T R32.3.1 does not work.

I still got the unexpected display after added the patch.

I actually replaced the new Image to original /boot/Image.
Then continue testing it.

The issue still could be duplicated easily.

Did I miss any?

Thank you,

Hi HuiW,

Yes, I guess you miss something. This patch is in nvgpu driver and nvgpu driver is a kernel module (.ko) file which is not in Image. It means you should replace nvgpu.ko under /lib/module/. instead of /boot/Image.

1 Like