[Bug]: [BUG] KWin 6.7.4-5.1 causes AMDGPU “Unsupported screen format RA24” on RX 9070 / Wayland; downgrade to 6.7.4-1.1 fixes it

### Kernel

6.18.42-1-cachyos-lts ~ ❯

### Desktop Environment / Window Manager

KDE Plasma

### CPU

AMD Ryzen 7 9800X3D 8-Core Processor

### GPU / Driver

AMD Radeon RX 9070

### Description

### Description

After updating CachyOS on 2026-08-17, KWin 6.7.4-5.1 causes repeated
AMDGPU DRM errors during Plasma Login Manager startup on Wayland.

Hardware:
- GPU: AMD Radeon RX 9070 (Navi 48, PCI ID 1002:7550)
- Display 1: AOC CU34G2XP, 3440x1440 @ 180 Hz, DisplayPort
- Display 2: HP 24f, 1920x1080 @ 74.97 Hz, DisplayPort
- Wayland
- Plasma Login Manager

Errors:

amdgpu 0000:03:00.0: [drm] *ERROR* Unsupported screen format RA24 little-endian (0x34324152)
kwin_wayland: Atomic modeset commit failed! Invalid argument
kwin_wayland: Applying output configuration failed!

### Regression

The issue started after the update on 2026-08-17:

kwin:
6.7.4-1.1 -> 6.7.4-5.1

Other packages updated at the same time included:
- mesa 26.1.6 -> 26.2.0
- linux-cachyos-bore 7.1.6 -> 7.1.8
- linux-firmware-amdgpu 20260622 -> 20260810

### Tests performed

1. linux-cachyos-bore 7.1.8 + kwin 6.7.4-5.1:
FAIL - RA24 / atomic modeset errors

2. linux-cachyos-lts 6.18.42 + kwin 6.7.4-5.1:
FAIL - same RA24 / atomic modeset errors

3. Forced 8 bpc for Plasma Login Manager:
FAIL - same RA24 errors

4. Downgraded ONLY KWin:
kwin 6.7.4-5.1 -> 6.7.4-1.1

RESULT:
SUCCESS

journalctl after reboot:
ZERO "RA24"
ZERO "Atomic modeset commit failed"
ZERO "Applying output configuration failed"

No monitor resolution or refresh rate was changed.
The AOC remains at 3440x1440 @ 180 Hz.
The session remains Wayland.

### Current workaround

KWin 6.7.4-1.1 works correctly.

I have temporarily added:

IgnorePkg = kwin

to /etc/pacman.conf to prevent 6.7.4-5.1 from being installed again.

### Conclusion

This appears to be a regression introduced by the CachyOS KWin
6.7.4-5.1 package/build, or a change contained in that revision,
because changing kernels did not fix the issue while downgrading
only KWin did.

### Steps to Reproduce

1. Boot CachyOS with KWin 6.7.4-5.1 using Plasma Login Manager on Wayland with both DisplayPort monitors connected (AMD Radeon RX 9070).

2. During Plasma Login Manager startup, check the current boot log:
sudo journalctl -b --no-pager | grep -Ei 'RA24|Atomic modeset commit failed|Applying output configuration failed'

3. The following errors are repeatedly produced:
amdgpu 0000:03:00.0: [drm] *ERROR* Unsupported screen format RA24 little-endian (0x34324152)
kwin_wayland: Atomic modeset commit failed! Invalid argument
kwin_wayland: Applying output configuration failed!

4. Downgrade only KWin from 6.7.4-5.1 to 6.7.4-1.1 and reboot, without changing the monitor configuration.

5. Run the same journalctl command again.

Result with KWin 6.7.4-1.1:
No RA24 errors, no atomic modeset failures and no output configuration failures.

### Expected Behavior

Plasma Login Manager / KWin Wayland should initialize both DisplayPort monitors without AMDGPU DRM format errors or atomic modeset failures.

The displays should keep their normal configuration:
- AOC CU34G2XP: 3440x1440 @ 180 Hz
- HP 24f: 1920x1080 @ 74.97 Hz
- Wayland session

This configuration worked correctly with KWin 6.7.4-1.1.

### Actual Behavior

After upgrading KWin from 6.7.4-1.1 to 6.7.4-5.1, Plasma Login Manager / KWin Wayland produces repeated DRM errors during startup:

amdgpu 0000:03:00.0: [drm] *ERROR* Unsupported screen format RA24 little-endian (0x34324152)

kwin_wayland: Atomic modeset commit failed! Invalid argument

kwin_wayland: Applying output configuration failed!

The issue also occurs with the CachyOS LTS kernel 6.18.42, so changing kernels does not resolve it.

Forcing 8 bpc also does not resolve it.

Downgrading ONLY KWin from 6.7.4-5.1 to 6.7.4-1.1 completely eliminates the problem. After rebooting with 6.7.4-1.1, the same journalctl check reports zero RA24, atomic modeset, or output configuration errors.

### Logs / Error Messages

//paste.cachyos.org/p/cf00974.log

### Additional Context

The issue appeared immediately after the CachyOS system update on 2026-08-17.

Relevant package changes during that update:
- kwin: 6.7.4-1.1 -> 6.7.4-5.1
- mesa: 26.1.6-1 -> 26.2.0-1
- linux-cachyos-bore: 7.1.6-1 -> 7.1.8-1
- linux-firmware-amdgpu: 20260622-1 -> 20260810-2

I tested the CachyOS LTS kernel 6.18.42 with KWin 6.7.4-5.1 and the RA24 / atomic modeset errors were still present.

I also tested forcing 8 bpc for Plasma Login Manager. This did not resolve the issue.

The decisive test was downgrading ONLY KWin:
6.7.4-5.1 -> 6.7.4-1.1

After rebooting, the errors completely disappeared:
- 0 RA24 errors
- 0 Atomic modeset commit failures
- 0 Applying output configuration failures

No monitor resolution, refresh rate or Wayland configuration was changed.

As a temporary workaround, I have added:
IgnorePkg = kwin
to /etc/pacman.conf.

I can retest a fixed/new KWin package if required.

Start by reproducing the issue with KWin 6.7.4-5.1 on the listed Wayland monitor setup and inspect the journalctl command and linked log. Compare the failing package with KWin 6.7.4-1.1; done means a corrected KWin package initializes both displays without RA24, atomic modeset, or output configuration errors.

Written by the indexing model from the issue text.