Created attachment 135115 [details]
Running the most recent code of agd5f/drm-next-4.15-dc leads to dmesg spam about vblank time outs (see attachment) and eventual full machine hangs when a 4k screen is connected to my RX560. The maching does not hang if only a 1080p screen is connected and it was running fine on commit d7205d5c0f2b09d900231de9026b97105c37d343.
Bisecting produced d050f8ed14494ea23c6e10b5893010946ef0f78b as the first bad commit, although on this commit the machine hangs immediatly after boot, so it might be a different issue.
(In reply to oschowa from comment #0)
> Bisecting produced d050f8ed14494ea23c6e10b5893010946ef0f78b as the first bad
> commit, although on this commit the machine hangs immediatly after boot, so
> it might be a different issue.
Since you can't test that commit for this problem, you should run
git bisect skip
for it, not "bad". Does the following commit boot and exhibit the problem?
OK, I took another look and the first commit which boots and exhibits the problem is 8459f633b65a81979b9b233ed0a91a9fdc98b7bd
I did a new kernel build with the most recent patches from agd5f/drm-next-4.15-dc applied and the issue persists. But then I manually reverted
c196cbe065da drm/amd/display: add dpms state to DC follow up
d050f8ed1449 drm/amd/display: add dpms state to DC
and the issue seems to be gone.
Fixed by the recent commits in agd5f/drm-fixes-4.15.