HDMI port monitor goes blank after 24 hours, but DisplayPort monitor remains operational

N

NSharp

New member
Strange problem with Ubuntu 18.04.3 LTS (with Linuxium's patched 5.3 kernel) -- I have all known "sleep" options turned off in Ubuntu for my NEO Z83-4U, but one monitor still goes blank after about 24 hours of running.

Privacy -> Screen Lock = Off
Power -> Blank screen = Never
Power -> Automatic suspend = Off
Devices -> Displays -> Mirror selected
Devices -> Displays -> Night Light = Off

The 2 monitors are the exact same model (ViewSonic VX2770SMH-LED). One monitor is being driven by a normal HDMI to HDMI cable, and the other is being driven by an active mini DisplayPort to HDMI cable. The DisplayPort monitor remained on for over 72 hours, but the HDMI monitor appears to go blank during that time.

I have multiple inputs to the HDMI monitor, and I can switch between a working Windows 10 source and the Ubuntu 18 source. But, the Ubuntu source is blank. It's almost like the monitor has gone to sleep, but it's not asleep or in power-save mode; I also do not get a "loss of signal" error message from the monitor (so it appears synced...). Lastly, I cannot get the monitor to become active again until I reboot Ubuntu. Changing the arrangement (joined versus mirrored) in Ubuntu does not bring the monitor back from being blank.

I see some dmesg logs that might be related:

[drm:intel_pipe_update_end [i915]] *ERROR* Atomic update failure on pipe A (start=14883587, end=14883588) time 528 us, min 1073, max 1079, scanline start 1052, end 1053

There are quite a few of these errors, but they vary in timing (as far as when they occur) and the "time" value in the log. The first occurrence of the log in dmesg is at almost 24 hours (85700.676386 seconds since boot).

Any ideas how to debug this issue? Is it a driver bug? I have plans to use the NEO Z83-4U as a digital signage device, but if I can't keep one of the monitors on at all times, that kind of defeats the purpose.

Thanks for any help,

Neil
 
@notrui2 and @ArcticWolf Thank you for your responses: yes, I am aware of the work @linuxium did on the kernel at boot time -- very thankful for it! :) I have installed his patched 5.3 kernel from the Canonical build server. I have no trouble booting up; my trouble starts approximately 24 hours into running, when the HDMI connection just disappears (goes blank).

I'd like to narrow down whether this is related to my particular setup or whether it is an edge case that is not covered in the patch -- or completely unrelated. It seems to be a long-term run case that may not have been observed previously. I wonder if the kernel or the i915 driver attempts to do communication similar to the way it does at boot time, every 24 hours or so? If so, could this be a case of the same issue at a different point in time? Not sure if the patch covers this case or not, if related.

I did find one interesting article related to the error in dmesg in my post: https://www.dedoimedo.com/computers/intel-microcode-atomic-update.html . I verified that the intel-microcode package (revision 0x411, dated 2019-04-23) is indeed installed on my system. So, I'll likely try his next suggestion to disable Panel Self-Refresh (PSR) in the i915 driver.
 
@NSharp: I'd start by adding

i915.enable_psr=0

to your kernel boot parameters as a lot of the related bug reports seem to indicate this fixes similar issues.
 
Well, despite still seeing the "Atomic update failure" messages in dmesg, my HDMI monitor is running as expected after nearly 26 hours now. The only change was the i915.enable_psr=0 module configuration change, in addition to the @linuxium 5.3 kernel. So, hopefully, this has resolved the issue.

Thanks for your help!
 
Well, after 6 days of running, the monitor blanked-out again -- even with the i915.enable_psr=0 workaround. Only a reboot recovers the screen.
 
I performed another experiment and restored the NEO Z83-4U back to the original Ubuntu 18.04.1 image from these forums. I prevented any system automatic updates and did not install any updates, aside from a handful of needed packages (e.g., paprefs). After 8 days, both monitors (HDMI and DisplayPort) are still running fine. I still see the "Atomic update failure" log in the dmesg logs, repeatedly; but, the system is stable.

So, the system is stable in 18.04.1. It is unstable in 18.04.3. The kernel patch provided by @linuxium fixes the boot-time HDMI issue (thanks!), but the long-term stability is still affected. Is MINIX looking into this video display issue at all in later versions of Ubuntu? Since the system will, by default, force the user to update the system, this is a pretty severe bug IMO.

My next test will likely be to hold back all kernel updates and just perform security patch updates for other packages. Then, after some time, I can update to the latest kernel release from this week.
 
I installed the latest ISO from MINIX (based on Ubuntu 18.04.3 LTS with kernel 5.0.0-32.34~18.04.2) and am currently testing this issue long-term. The new kernel should contain the HDMI fixes mentioned elsewhere in these forums.
 
You must log in or register to reply here.
Back
Top