Created attachment 281193 [details] lsmod I am not 100% sure whether this bug falls under i915, pwm-lpss, or other. I have an ASUS Transformer Mini T103HA with an Intel Cherry Trail chip. Distro is Slackware Current with self-built kernel. I am running kernel 4.20.4, vanilla, except I added a patch to get the sound codec (Conexant cx2072x) working. I took the .config from Fedora Rawhide and pared it down a bit to suit my needs. lspci -vvv, dmesg and lsmod attached. This seems to especially happen after my machine has been suspended and resumed twice or more, or perhaps when my machine has been suspended in S3 for a while, and subsequently resumed. The backlight has a beat frequency best described as a flicker. This is especially visible when grey colours are shown on screen. In fact when grey colours are shown on screen, you can sometimes even see some of the window behind the window that is active, even with compositing disabled. This is quite strange and I've never seen anything quite like it. This issue affects Xorg (xorg-server 1.20.3) and wayland (weston 5.0.0). Rebooting resolves the issue. I *think* the issue seems to go away after letting my machine sit with the screen on for a while, but I could be wrong and it could just be my eyes getting used to it. Not 100% sure. I get the feeling the pwm frequency is getting thrown off somehow. I tried the following: # intel_reg_read 0xC8254 the output was: 0xC8254 : 0x0 Probably because this SoC isn't using the i915 registers for its backlight and is using its own pwm controller. Just a guess.
Created attachment 281195 [details] lspci -vvv
Created attachment 281197 [details] dmesg
By way of update, Hans de Goede from Red Hat suggested that I test this under FC29 (with kernel 4.18.16) and try updating to the 4.20.x kernel. I *cannot* reproduce the flicker issue under 4.18.16. I was not able to try under 4.20.x under Fedora (only self-built kernel under Slackware which is where I initially experienced this) but I was able to try with a RawHide build using kernel 5.0-rc6, and I *can* reproduce there. I'm wondering at this point if this is a possible regression. I did notice a bit *more* of the odd transparency/artifacting issue as described above under 4.18.x, but absolutely none of the flicker upon suspend/resume.
I can no longer reproduce this bug under 5.3.4.
Never mind - I can reproduce. It just took longer to happen the first time.