What "gk20a 57000000.gpu: Timeout detected @ gk20a_pmu_disable_elpg+0x210/0x2f0" means?


After rebooting I found in the dmesg log the following lines:

gk20a 57000000.gpu: Timeout detected @ gk20a_pmu_disable_elpg+0x210/0x2f0
gk20a 57000000.gpu: gk20a_pmu_disable_elpg: ELPG_DISALLOW_ACK failed

Should I be concerned about those lines?

Then when I run my cuda application I got again the line:

gk20a 57000000.gpu: Timeout detected @ gk20a_pmu_disable_elpg+0x210/0x2f0



Does this one cause any gpu related issue? I mean rendering, cuda or something else.

Please describe more about your environment. (custom carrier board/devkit? How to reproduce this issue?)

Hi Wayne,

Thanks for replying. No, it does not seem to cause any gpu issue, but I would love to know why it happened.

Custom carrier board, 28.1 devkit and monolithic kernel.

I don’t know the details of this, but for your curiosity I may add that I suppose that GPU was told to disable Engine Level clock gating (elcg, probably for switching to another clock gating mode), it was expected to acknowledge this with an interrupt when done, but this hasn’t happened before the timeout, so this trace.

If all CUDA stuff works fine after boot, it may not be an issue.
On the other hand, considering you have a custom board and custom kernel, you may investigate this.
Maybe checking the original R28.1 kernel with DTB adjustments for your custom board could rule out if this is board or kernel related.

Someone with better knowledge may comment further.