Bug 105083

Please attach the output of dmesg captured while the problem is occurring, preferably with amdgpu.dc_log=1 .

(In reply to denisgolovan from comment #0)
> Hi
> 
> Upgrading to 4.15.2 with amdgpu dc results in blinking of color temperature
> when redshift application started. Kernel 4.13.13 worked fine.
> 
> See https://github.com/jonls/redshift/issues/221#issuecomment-365376437 for
> details

Did you enable AMDGPU DC by changing kernel boot parameters?

(In reply to Michel Dänzer from comment #1)
> Please attach the output of dmesg captured while the problem is occurring,
> preferably with amdgpu.dc_log=1 .

I tried to capture some dmesg log, but booting with suggested kernel line amdgpu.dc_log=1 but no output was produced at flickering. Than I realized that I have to enable amdgpu.dc in order to use it's logging :) and the issue disappeared. So far after a hour I don't see any flickering anymore.

Adding amdgpu.dc=1 amdgpu.dc_log=1 indeed make flickering much less frequent.
However, I can still reproduce them when turning redshift on and off.

No logs in dmesg and messages with amdgpu appear though.
Any hints?

How bad is the flicker? Would you be able to take a video showing the flicker and post it (youtube or anywhere, really)?

(In reply to Harry Wentland from comment #4)
> How bad is the flicker? Would you be able to take a video showing the
> flicker and post it (youtube or anywhere, really)?

Video Card: RX 480

As I am suffering as well from this problem, I have uploaded a video to youtube.

https://youtu.be/fMNhPH6QG00

the flicker happens periodically as Redshift adjusts the color temperature over time.

(In reply to Allen from comment #5)

> the flicker happens periodically as Redshift adjusts the color temperature
> over time.

Yes. Seems pretty similar.
However it happens also periodically when redshift is active.

(In reply to denisgolovan from comment #6)
> (In reply to Allen from comment #5)
> 
> > the flicker happens periodically as Redshift adjusts the color temperature
> > over time.
> 
> Yes. Seems pretty similar.
> However it happens also periodically when redshift is active.

I agree, this is the flickering I'm seeing, and it does happen occasionally when redshift is running even when it's not adjusting the color temperature.  It's much more frequent during the transition times when it is adjusting.

If I disable either redshift or amdgpu DC, I don't see any flickering.

Thanks for the video. It helps us understand what you're seeing.

I get random blue color screen flickering perhaps once every few minutes on each monitor with redshift and a RX 580 and a 3x4k 60Hz setup with kernel 4.15.3 and amdgpu.dc=1. My initial response was that amdgpu.dc is broken so I reverted back to 4.14.x and used that until I found this bug and realized it's the redshift that's badly handled / problematic with the amdgpu.dc code.

This isn't really a redshift bug, just the way it behaves with this new code since I've used redshift since forever on numerous configurations and this is a new problem.

The random blue flickering with redshift seems to randomly appear on one of the monitors, not all or two at once. All do not blink the same time. I can't really see any pattern either. Every few minutes one of the monitors will blink blue. Sometimes none of them blink blue for a long time.

Everything seems fine so far without redshift apart from the blue color spectrum now violently smashing my eyes. My eyes, my eyes! I'll test if redshift is OK with transition = 0 later, it could be a workaround if it is the gradual adjustment of gamma/color that causes the flickering.

We have some patches that rework our gamma programming in our staging trees. Would you be able to try amd-staging-drm-next or drm-next-4.17-wip from https://cgit.freedesktop.org/~agd5f/linux/?h=drm-next-4.17-wip ?

(In reply to Harry Wentland from comment #10)
> We have some patches that rework our gamma programming in our staging trees.
> Would you be able to try amd-staging-drm-next or drm-next-4.17-wip from
> https://cgit.freedesktop.org/~agd5f/linux/?h=drm-next-4.17-wip ?

I've just built drm-next-4.17-wip + amdgpu.dc=1.
Well, it does not blink however it has another regression - color temperature/gamma is not applied to all window controls. Some areas just stay not affected as if gamma change specifically disabled for them.

Disabling amdgpu.dc under drm-next-4.17-wip works fine though.

(In reply to denisgolovan from comment #11)
> (In reply to Harry Wentland from comment #10)
> Some areas just
> stay not affected as if gamma change specifically disabled for them.

Do you have a picture?

(In reply to Harry Wentland from comment #13)
> (In reply to denisgolovan from comment #11)
> > (In reply to Harry Wentland from comment #10)
> > Some areas just
> > stay not affected as if gamma change specifically disabled for them.
> 
> Do you have a picture?

See https://www.youtube.com/watch?v=6MQrgemv-yQ
White color on github page is completely unaffected when I switch Redshift on/off.

Any news on that?

The ticket is on our backlog and we'll try to get to it in a week or so.

Experiencing this since 4.15 (not sure which minor release). Still happening on 4.16.0. With and without amdgpu=1, using RX480:

01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/580] (rev c7)

A fix should be in drm-next-4.17 of Alex's git repo at https://cgit.freedesktop.org/~agd5f/linux/?h=drm-next-4.17

That fix should also make it into 4.16 stable.

White color fix with redshift/nightshift should also be in that branch.

Cannot confirm your fix.

amdgpu.dc=1 still blinks. 
Using old engine (disabling dc) works fine.
Tested under vanilla Linux v4.16.8.

White color regression is fixed in both modes.

Since this bug remains open for some reason I'll report the following: redshift appears to work just fine with both RX 560 & RX 470 and:

- kernel 4.17.0-1 (rx 580) / 4.18.0-0.rc0.git2 (rx 470)
- X.Org 1.19.6
- mesa 18.0.2
- xfce xfwm4 4.12.4
- compton-0.1-0.1.beta3

kernel commmand line has amdgpu.dc=1 for some reason (default now anyway). I assume both kernels would work fine on both cards, that's just what happens to be running right now.

I suspect this bug can be closed but it's probably fair to wait for them russians to confirm.

Tested under 4.17.2.
Works fine with amdgpu.dc=1