[BUG] linux-cachyos 7.1.6: Plasma login screen freezes and turns black due to amdgpu pageflip timeouts; LTS kernel works
bug
### Pre-flight checklist
- [x] I have searched existing issues and this is not a duplicate.
- [x] I have read the [Contributing Guidelines](https://github.com/CachyOS/linux-cachyos/blob/master/CONTRIBUTING.md).
- [x] I have verified the issue is reproducible with the latest available CachyOS kernel.
- [x] I have tried to reproduce the issue on Arch Linux's `linux` kernel.
### Upstream / vanilla kernel check
I have not tested with a vanilla/upstream kernel
### Kernel variant
linux-cachyos (EEVDF, Clang)
### System information (cachyos-bugreport.sh)
```text
The bug report was generated while running the working CachyOS LTS kernel,
because the regular linux-cachyos kernel does not reach a usable graphical
session.
CachyOS bug report:
https://paste.cachyos.org/p/b5e77d9.log
```
### Manual system information (if cachyos-bugreport.sh is unavailable)
```text
OS: CachyOS x86_64
Device: MSI Alpha 17 C7VG, revision 1.0
Bootloader: Limine
Filesystem: Btrfs
CPU: AMD Ryzen 9 7945HX
Hybrid graphics:
- AMD Radeon 610M / Raphael integrated GPU, amdgpu driver
- NVIDIA GeForce RTX 4070 Max-Q / Laptop GPU
- NVIDIA Open Kernel Module 610.57.04
Failing kernel:
- linux-cachyos 7.1.6-1
- linux-cachyos-nvidia-open 7.1.6-1
Working kernel:
- linux-cachyos-lts 6.18.42-1
- linux-cachyos-lts-nvidia-open 6.18.42-1
Desktop:
- KDE Plasma 6.7.3
- KWin 6.7.3
- Wayland
- plasma-login-manager 6.7.3-2
- Mesa 26.1.6-1
- Plymouth 26.134.222-2.1
Display:
- Internal 2560x1440 laptop panel
- Configured for up to 240 Hz
- The internal eDP panel is handled by the AMD integrated GPU
Architecture/repositories:
- CachyOS znver4 repositories enabled
No SCX scheduler was intentionally configured.
```
### Bug description
After booting the regular linux-cachyos kernel, the Plasma Login Manager
appears initially, then freezes and the internal laptop display turns black.
The system does not recover and must be powered off forcibly.
The problem is reproducible with linux-cachyos 7.1.6-1. Booting the same
installation with linux-cachyos-lts 6.18.42-1 works normally. KDE Plasma,
KWin, Mesa, the login manager and the GPU userspace driver packages are the
same in both cases.
The failed boot used the normal Limine boot entry and was not a Btrfs snapshot
boot.
The journal shows that Plymouth terminates normally, Plasma Login Manager
starts successfully, and the Wayland greeter connects. Immediately afterwards,
KWin repeatedly reports pageflip timeouts in the amdgpu kernel driver.
### Steps to reproduce
1. Fully update CachyOS.
2. Reboot the MSI Alpha 17 C7VG.
3. Select the normal linux-cachyos 7.1.6-1 entry in Limine.
4. Wait for the Plasma Login Manager Wayland greeter.
5. The login screen appears, freezes within a few seconds and the display
turns black.
6. The system must be powered off using the physical power button.
7. Boot linux-cachyos-lts 6.18.42-1 instead.
8. The login screen and KDE Plasma desktop work normally with the LTS kernel.
### Expected behavior
The Plasma Login Manager should remain responsive and allow the user to log
into the KDE Plasma Wayland session.
### Actual behavior
The Plasma Login Manager starts, but KWin's DRM display output stops updating.
The internal display freezes and then sometimes becomes black. Sometimes it works a few seconds before freezing and allows typing. One time login was successful before freezing, but could not reproduce. Freeze time is not always the same, but seems somewhat random.
The journal continuously reports:
Pageflip timed out! This is a bug in the amdgpu kernel driver
A forced power-off is required.
### Logs / stack traces
```text
Aug 05 11:55:49 plasmalogin[1081]: Greeter starting...
Aug 05 11:55:49 plasmalogin-helper[1083]: pam_unix(plasmalogin-greeter:session): session opened for user plasmalogin(uid=959) by (uid=0)
Aug 05 11:55:49 plasmalogin-helper[1083]: Starting Wayland user session: "/usr/share/plasmalogin/scripts/wayland-session" "/usr/bin/startplasma-login-wayland"
Aug 05 11:55:49 plasmalogin[1081]: Greeter session started successfully
Aug 05 11:55:50 kwin_wayland[1124]: No backend specified, automatically choosing drm
Aug 05 11:55:50 kwin_wayland[1124]: Accepting client connections on sockets: QList("wayland-0")
Aug 05 11:55:51 plasma-login-greeter[1184]: Connected to the daemon.
Aug 05 11:55:54 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Aug 05 11:55:55 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Aug 05 11:55:57 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Aug 05 11:55:58 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Aug 05 11:55:59 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Aug 05 11:56:00 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Aug 05 11:56:01 kwin_wayland[1124]: Pageflip timed out! This is a bug in the amdgpu kernel driver
Relevant amdgpu initialization:
amdgpu 0000:06:00.0: [drm] Display Core v3.2.378 initialized on DCN 3.1.5
amdgpu 0000:06:00.0: [drm] DMUB hardware initialized: version=0x05003000
amdgpu 0000:06:00.0: [drm] eDP-2: PSR support 1, DC PSR ver 0,
sink PSR ver 1, DPCD caps 0x2a, su_y_granularity 0
[drm] Initialized amdgpu 3.64.0 for 0000:06:00.0
fbcon: amdgpudrmfb (fb0) is primary device
```
### Additional system information
- Plymouth terminates successfully before Plasma Login Manager starts.
- Plasma Login Manager reports "Greeter session started successfully".
- No NVIDIA API mismatch is present; both the kernel module and nvidia-utils
use version 610.57.04.
- The NVIDIA GPU has no connected CRTC because this is a hybrid-graphics
laptop and the internal panel is attached to the AMD integrated GPU.
- linux-cachyos-lts 6.18.42-1 works with the same userspace packages.
- I have not yet applied amdgpu.dcdebugmask or other amdgpu workaround
parameters.
- I have not yet tested an unmodified Arch/upstream kernel.
- Full journal and kernel journal from the failed boot are attached.
Possibly unrelated consequence: after the forced power-off, the first boot
with the LTS kernel sometimes has no Wi-Fi. A second normal LTS reboot restores
Wi-Fi.
[failed-boot-kernel.txt](https://github.com/user-attachments/files/30740206/failed-boot-kernel.txt)
[failed-boot-full.txt](https://github.com/user-attachments/files/30740205/failed-boot-full.txt)
[cachyos-bugreport.log](https://github.com/user-attachments/files/30740207/cachyos-bugreport.log)A
9 条评论