Seems there are some positive shifts towards a solution after updating to Plasma 2.1.5
Operating System: Manjaro Linux
KDE Plasma Version: 6.1.5
KDE Frameworks Version: 6.6.0
Qt Version: 6.7.2
Kernel Version: 6.6.54-2-MANJARO (64-bit)
Graphics Platform: Wayland
Processors: 16 × AMD Ryzen 9 7940HS w/ Radeon 780M Graphics
Memory: 30.5 GiB of RAM
Graphics Processor: AMD Radeon 780M
Manufacturer: Framework
Product Name: Laptop 16 (AMD Ryzen 7040 Series)
System Version: AJ
Along the way my DP input on monitor died, so I’ve switch to connecting it via HDMI-to-USBC.
Post Plasma Update, I had a couple casual games running via gamescope… and I thought I’d see what hapens if I simply unplugged the external monitor for the side port and direct connected it to the dGPU.
I was plesantly surprised when all seemed to be well… no “invisible lines” all areas clickable… things were looking up.
Then I figured that I should reboot just to see what happens then… and SDDM only displaying on the laptop display (as expected having not updated /usr/share/sddm/scripts/Xsetup), I logged in, both monitors were displaying within the wayland session, launched the same two casual games and all seemed normal… no “invisible lines” or unclickable areas.
Then I started rereading through this thread to see what other tests might be relevant to retry and launched a full screen game (Grim Dawn). Things looked well once again with no invisible lines.
However, this is where things went off the rails.
I noticed the refresh rate was lower than normal (see Worse gaming (FPS) performance after upgrading PPD 0.21-1 to PPD 0.23-1) and adjusted the PPD slider to balanced… accessng the slider by pressing Meta/Win key to expose the system tray and access the slider
At some point after pressing the Meta key and either before/after adjusting the slider… I had a slight freeze which I think might have been a plasma or xwayland crash/recovery… and had to close the game with Steam as the game stopped presenting visual changes.
Relaunched the game just to confirm the frame rate increased, which it did, but when I closed the game via it’s internal enu/option… the PC froze so hard that even REISUB would not save me.
So I’m not sure if adjusting the PPD slider with the full screen game running caused the crash… but I was able to do that without issues while games were running in a gamescope window before moving my display cable to the dGPU.
Analyzing the journal afterwards, it looks like 10 minutes after boot (I suspect after i launched closed the casual games and opened the fullscreen game), I got the following in
journalctl -p 3 -b -1 -x
Oct 17 11:41:59 AMD-Ryzen9-7940HS kwin_wayland[996]: kwin_wayland_drm: Pageflip timed out! This is a kernel bug
Oct 17 11:42:16 AMD-Ryzen9-7940HS kernel: [drm:dm_plane_helper_prepare_fb [amdgpu]] *ERROR* Failed to pin framebuffer with error -22
That last line repeated hundred of times until xwayland coredump … (I suspect when game/system crash/froze)
Oct 17 11:46:13 AMD-Ryzen9-7940HS systemd-coredump[67568]: [🡕] Process 1092 (Xwayland) of user 1000 dumped core.
I figure/hope that my still being on lts kernel 6.6, might explain either of these journal entries… and that when lts kernel 6.12 comes out that might change things for the better… however, one is related to kwin_wayland… so at least part of he issue appears to be with KDE.
Now I have to figure out what damage happened in my home folder during the crash…
$journalctl -p 3 -b 0
Oct 17 11:49:12 AMD-Ryzen9-7940HS kernel: EXT4-fs error (device nvme1n1p3): ext4_orphan_get:1424: comm mount: bad orphan inode 33297853
Oct 17 11:49:12 AMD-Ryzen9-7940HS kernel: ext4_test_bit(bit=5564, block=133169168) = 0
$ lsblk -f
NAME FSTYPE FSVER LABEL UUID FSAVAIL FSUSE% MOUNTPOINTS
nvme1n1
├─nvme1n1p1 vfat FAT32 EFI SYSTEM 1FC6-DE58 510.7M 0% /boot/efi
├─nvme1n1p2 ext4 1.0 Manjaro 0e2d6c29-a2a6-4fd2-a472-78713cd43975 67.1G 26% /
├─nvme1n1p3 ext4 1.0 Home 2c1b4c48-6cd6-4f81-af2b-42f766665d5f 89.1G 83% /home
└─nvme1n1p4 swap 1 Swap e819436d-6704-4578-a97f-2000170f14af [SWAP]
nvme0n1
└─nvme0n1p1 ext4 1.0 Backup 016ce75b-2aeb-458b-a403-94b06ac57be8 178.1G 76% /data/sn740
I’ll try unmount /Home in a TTY post logout… but I suspect I’ll need a live USB key to boot and then check the drive from there.
EDIT:
Log out, switch to tty, umount and e2fschk only prompted concerning optimization of inodes which I let it do and rebooted.
I got curious about what file the inode from the journal was referring to, and I think only impacting a firefox cache file is pretty lucky/fortunate
$ find . -inum 33297853
./.cache/mozilla/firefox/bzlnfjz7.default-release/cache2/entries/18C0A1412751BC8FB583B3AA48924987874873F1
With the cable moved back tot he side port (iGPU) I launched Grim Dawn fullscreen again, noticed that the frame rate was 100 instead of 50 (1/2 fps while connected to dGPU?), closed the game without issue, and confirmed no kwin/drm/coredump errors in the journal… yah, gonna keep the connection to the iGPU until lts kernel 6.12 (and KDE Plasma 6.2+) later this year and then try again.
EDIT 2:
Per my revised Worse gaming (FPS) performance after upgrading PPD 0.21-1 to PPD 0.23-1 thread, the issue with lower game performance was upgrading PPD to 0.23-1 and downgrading it back to 0.21-1 solved the performance issues.
On the surface it seems many of my graphical issues (like “invisible lines” clipping video output) have been resolved… but underlying amdgpu (and/or kwin_wayland) issues remain. I’m hopeful the next 6.12 LTS kernel and all it’s combined AMD fixes/features added post LTS 6.6 will help solve these.
kwin_wayland[996]: kwin_wayland_drm: Pageflip timed out! This is a kernel bug
kernel: [drm:dm_plane_helper_prepare_fb [amdgpu]] *ERROR* Failed to pin framebuffer with error -22