Project

General

Profile

Actions

Bug #315

open

Xorg/Wayland show black screen when system memory is 4GB or higher

Added by Master Geek over 3 years ago. Updated over 3 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
Start date:
08/02/2021
Due date:
% Done:

0%

Estimated time:
Affected versions:
Needs backport to:
Affected hardware:
Affected OS:

Description

Board: ASROCK G41M-VS3 R2.0
CB Revision: coreboot-4.14-1258-gfbc46a3bfb

When this board has one memory stick with 2GB of ram, or two memory sticks with 1GB and 2GB (3GB total) there's no issue with display; Xorg and wayland (using weston) work as expected. But when there's 4GB or maybe more (two 2GB sticks only available, haven't tried a single 4GB stick) the system works as expected up to just before starting Xorg or wayland, the screen goes black and nothing else happens. Returning to a tty by means of Ctrl-Alt-# 'ing to some other tty doesn't bring display back, neither does exitting/killing Xorg or wayland. Not even forcefully removing and reloading the relevant kernel modules (i915, drm) does anything to bring the display back, display will only comeback with a reboot.

When the screen goes black there's no system freeze of any kind, and the kernel complains with a single line:
"i915 0000:00:02.0: [drm] ERROR CPU pipe A FIFO underrun"
that isn't very helpful for debugging imo.

Attached dmesg loglevel=9 logs of stock bios, coreboot with 4GB (coreboot_bad) and with 2GB (coreboot_good), cbmem console log of coreboot with 2GB (cbmem_good).


Files

cb_config.txt (16.3 KB) cb_config.txt Coreboot config file Master Geek, 08/02/2021 06:32 PM
cbmem_good.log (31.3 KB) cbmem_good.log cbmem output working (2GB) Master Geek, 08/02/2021 06:32 PM
dmesg_coreboot_good.log (457 KB) dmesg_coreboot_good.log dmesg output working (2GB) Master Geek, 08/02/2021 06:32 PM
dmesg_coreboot_bad.log (443 KB) dmesg_coreboot_bad.log dmesg output not working (4GB) Master Geek, 08/02/2021 06:32 PM
dmesg_stock_bios.log (315 KB) dmesg_stock_bios.log dmesg output stock bios (4GB) Master Geek, 08/02/2021 06:32 PM
cbmem_good.log (36.6 KB) cbmem_good.log cbmem output working (2GB) Master Geek, 08/02/2021 06:55 PM
cbmem_bad.log (37 KB) cbmem_bad.log cbmem output not working (4GB) Master Geek, 08/02/2021 06:55 PM
Actions #2

Updated by Master Geek over 3 years ago

First cbmem_good.log file was made with a build with console log level set to 6 (INFO). The cbmem logs below are from a build with log level 7 (DEBUG).

Actions #3

Updated by Master Geek over 3 years ago

After removing some jumper connectors on the board I was able to boot and start Xorg/Wayland while using 4GB of ram (2 sticks of 2GB each). Have yet to try and relocate the jumpers in the same positions as they were to reproduce the issue but I think that's a quirk of this specific board of mine. The jumpers had to do with something called 'EuP', I think it has to do with power settings but I'm not sure.

There's one for globally enabling/disabling it, and other two for selectively enabling it for Audio and LAN; when I got it booting I noticed audio just stopped working. Linux detected the soundcard as always and it wasn't muted nor was the volume low but there was genuinely no output through the stereo jack of the case. Replacing the jumper for the audio got it enabled.

I think this issue at the very least should have a lower severity/priority now.

Actions #4

Updated by Master Geek over 3 years ago

Honoring PaulePanther's request I document here that it seems 'EuP' (Energy Using Product) is related to a power saving, power efficiency regulation of sorts. As per this document:

"EuP, stands for Energy Using Product, was a provision regulated by European Union to define the power consumption for the completed system. According to EuP, the total AC power of the completed system shall be under 1.00W in off mode condition. To meet EuP standard, an EuP ready motherboard and an EuP ready power supply are required. According to Intel’s suggestion, the EuP ready power supply must meet the standard of 5v standby power efficiency is higher than 50% under 100 mA current consumption."

So it kind of makes sense for it to play bad with coreboot somehow...

Actions

Also available in: Atom PDF