I suspect this isn't an amdgpu bug but rather an ftrace bug exposed by amdgpu.

Two ideas.

1) On a "bad" kernel can you compile without ftrace support and see if it helps the issue. Around that timeframe I know there was a change to ftrace that fixes it getting hung up at modules with lots of trace points (like amdgpu).

2) Could you keep going on your bisect to get down to a specific commit?

Reply via email to