[drm:amdgpu_ctx_mgr_entity_fini [amdgpu]] *ERROR* ctx 00000000189dec14 is still alive

[孙南勇]

Hi All,
    Sometime dmesg show this log:
[10407.720988] [drm:amdgpu_ctx_mgr_entity_fini [amdgpu]] *ERROR* ctx 00000000189dec14 is still alive
[10407.721054] [drm:amdgpu_ctx_mgr_entity_fini [amdgpu]] *ERROR* ctx 00000000189dec14 is still alive
[10407.721120] [drm:amdgpu_ctx_mgr_entity_fini [amdgpu]] *ERROR* ctx 00000000189dec14 is still alive
[10407.721187] [drm:amdgpu_ctx_mgr_entity_fini [amdgpu]] *ERROR* ctx 00000000189dec14 is still alive
[10407.721418] [drm:amdgpu_ctx_mgr_fini [amdgpu]] *ERROR* ctx 00000000189dec14 is still alive



What are the side effects?
Does this error will occur GPU hung(such as gfx timeout)?
How to fix this?


My kernel : 4.19.36
Mesa: 18.3.6


And I find that in amdgpu_ctx.c, the function:amdgpu_ctx_mgr_entity_flush doesn't check the 
ctx->refcount(the check is stopped in this patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=48ad368a8a3ab2fd3c2bc2ccccc6e29b1acda1bb) , 
 why amdgpu_ctx_mgr_entity_fini still need check ctx->refcount?
Can i stop check ctx->refcount in  amdgpu_ctx_mgr_entity_fini ?


Thanks!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20190805/763f1f17/attachment.html>