Title: [linux-cachyos-rc 7.2.rc3-1] amdgpu gfx hang on Granite Ridge iGPU: gfxhub page fault -> ring reset failed -> MODE2 reset

### 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-rc

### System information (cachyos-bugreport.sh)

```text
## Summary

Single GPU hang during normal desktop use (KDE Plasma Wayland, typing in a
GPU-accelerated terminal). The iGPU raised gfxhub page faults from the
compositor's command stream, the gfx ring timed out, the per-ring reset
FAILED, and the driver escalated to a MODE2 full GPU reset, which succeeded
in ~2s. Applications holding GL contexts lost them (kitty crashed, kwin
recovered on its own). No persistent damage.

## Hardware

- CPU: AMD Ryzen 9 9950X3D
- Memory: 96GB ddr5
- GPU driving the desktop: Granite Ridge iGPU [1002:13c0 rev c9], GC 10.3.6, 2048M VRAM
- Also present: NVIDIA RTX 5090 [10de:2b85] with nvidia-open 7.2.rc3 modules (not compositing)

## Software

- Kernel: linux-cachyos-rc 7.2.rc3-1 (uname: 7.2.0-rc3-1-cachyos-rc)
- mesa 3:26.1.5-1, vulkan-radeon 3:26.1.5-1
- KDE Plasma Wayland session, kwin 6.7.3

## dmesg (trimmed; full excerpt attached)

amdgpu 0000:7a:00.0: [gfxhub] page fault (src_id:0 ring:24 vmid:4 pasid:24)
amdgpu 0000:7a:00.0: Process kwin_wayland pid 7860 thread kwin_wayla:cs0 pid 7908
amdgpu 0000:7a:00.0: in page starting at address 0x0000004380003000 from client 0x1b (UTCL2)
amdgpu 0000:7a:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x00401031
amdgpu 0000:7a:00.0: Faulty UTCL2 client ID: TCP (0x8)
amdgpu 0000:7a:00.0: MORE_FAULTS: 0x1
amdgpu 0000:7a:00.0: PERMISSION_FAULTS: 0x3
amdgpu 0000:7a:00.0: RW: 0x0
(3 more faults, addresses 0x4380001000-0x4380003000)
amdgpu 0000:7a:00.0: [drm] AMDGPU device coredump file has been created
amdgpu 0000:7a:00.0: ring gfx_0.0.0 timeout, signaled seq=3872273, emitted seq=3872276
amdgpu 0000:7a:00.0: Process kitty pid 11946 thread kitty:cs0 pid 11947
amdgpu 0000:7a:00.0: Starting gfx_0.0.0 ring reset
amdgpu 0000:7a:00.0: Ring gfx_0.0.0 reset failed
amdgpu 0000:7a:00.0: GPU reset begin!. Source: 1
amdgpu 0000:7a:00.0: MODE2 reset
amdgpu 0000:7a:00.0: GPU reset(1) succeeded!
amdgpu 0000:7a:00.0: [drm] device wedged, but no recovery needed

The amdgpu devcoredump (5.4 MB) was captured and is attached (zipped).

## Notes / caveats

- First occurrence ever on this machine; not reproducible so far.
- This machine has run the 7.2-rc series (rc1, rc2, rc3) since early July
with zero GPU faults until this event.
- Possible confounder: mesa was upgraded 26.1.4 -> 26.1.5 about 2 hours
before the hang, mid-session - the compositor still had 26.1.4 mapped in
memory while newer processes had loaded 26.1.5. The session had not been
restarted since the upgrade.
- A suspend/resume cycle happened ~7h before the hang in the same boot.
- I will report back if it recurs after a clean reboot on consistent
versions; I can also boot linux-cachyos 7.1.3 (stable) for triage if
useful.
```

### Manual system information (if cachyos-bugreport.sh is unavailable)

```text

```

### Bug description

amdgpu gfx hang on Granite Ridge iGPU: gfxhub page fault -> ring reset failed -> MODE2 reset

### Steps to reproduce

Not reproducible

### Expected behavior

Not reproducible

### Actual behavior

[linux-cachyos-rc 7.2.rc3-1] amdgpu gfx hang on Granite Ridge iGPU: gfxhub page fault -> ring reset failed -> MODE2 reset

### Logs / stack traces

```text
[sysinfo.txt](https://github.com/user-attachments/files/30105315/sysinfo.txt)
[kernel-log-excerpt.txt](https://github.com/user-attachments/files/30105314/kernel-log-excerpt.txt)
[devcoredump.txt](https://github.com/user-attachments/files/30105313/devcoredump.txt)
```

### Additional system information

_No response_