[BUG] [Regression] dcn32_program_compbuf_size REG_WAIT timeout persists on DCN 3.2 (RX 7900 XTX) after #810 fix — HDMI 4K@120Hz mode not exposed

[DegenerationDX]

Pre-flight checklist

• I have searched existing issues and this is not a duplicate.
• I have read the Contributing Guidelines.
• I have verified the issue is reproducible with the latest available CachyOS kernel.
• 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)

CPU: AMD Ryzen 7 7800X3D
- dGPU: AMD Radeon RX 7900 XTX (Navi 31, PCI 0000:03:00.0, PCI ID 0x1002:0x744c), 24GB VRAM, DCN 3.2
- iGPU: AMD Raphael (PCI 0000:18:00.0), also DCN 3.2 family, present but not driving the affected display
- RAM: 64GB
- Display: Samsung Q85A (2021 QLED, HDMI 2.1 capable) connected via HDMI-A-1
- Display server: KDE Plasma 6.7.4, KWin on Wayland

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

Bug description

This is the DCN 3.2 counterpart of #810, which was fixed and closed on 2026-07-07 for dcn31_program_compbuf_size (DCN 3.1, drivers/gpu/drm/amd/display/dc/hubbub/dcn31/dcn31_hubbub.c). That fix replaced the four open-coded REG_WAIT(DCHUBBUB_DETx_CTRL, ...) calls with dcn31_wait_for_det_apply() and was merged upstream via the amd-gfx patch "drm/amd/display: fix compressed buffer config routine waiting time".
However, that patch only touched dcn31_hubbub.c. The equivalent function for DCN 3.2 hardware, dcn32_program_compbuf_size() in drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hubbub.c, still contains the original unfixed code:

if (compbuf_size_segments > hubbub2->compbuf_size_segments) {
    REG_WAIT(DCHUBBUB_DET0_CTRL, DET0_SIZE_CURRENT, hubbub2->det0_size, 1, 100);
    REG_WAIT(DCHUBBUB_DET1_CTRL, DET1_SIZE_CURRENT, hubbub2->det1_size, 1, 100);
    REG_WAIT(DCHUBBUB_DET2_CTRL, DET2_SIZE_CURRENT, hubbub2->det2_size, 1, 100);
    REG_WAIT(DCHUBBUB_DET3_CTRL, DET3_SIZE_CURRENT, hubbub2->det3_size, 1, 100);
}


### Steps to reproduce

1.Boot CachyOS with kernel 7.2.0-1 or later (regression first appeared after upgrading from 7.1.8-1 to 7.2.0-1) on an RX 7900 XTX.
2.Connect a 4K120/HDMI 2.1 capable display (Samsung Q85A in this case) via HDMI.
3.Check dmesg for amdgpu/DC errors during boot.
4.Check available modes with kscreen-doctor -o or cat /sys/class/drm/card*-HDMI-A-*/modes.

### Expected behavior

The amdgpu driver should program the compressed buffer (DET) sizes without a register wait timeout, and the full set of modes the display's EDID advertises (including 3840x2160@120Hz / 4096x2160@120Hz) should be available to KWin/KScreen.

### Actual behavior

amdgpu 0000:03:00.0: [drm] REG_WAIT timeout 1us * 100 tries - dcn32_program_compbuf_size line:148

(Also seen at line:147 in other reports.) The 4K@120Hz / 4K@100Hz modes are absent from the connector's mode list; the TV is capped at 3840x2160@60Hz / 4096x2160@60Hz despite EDID support for FRL-based 120Hz timings.

7.1.8-1: not affected. kscreen-doctor -o correctly lists 4096x2160@120.00, 3840x2160@120.00 (active, marked *), 3840x2160@119.88, 3840x2160@100.00, etc.
7.2.0-1: regression appears immediately after upgrading and rebooting. 4K modes capped at 60Hz, no 100/120Hz variants listed at all.
7.2.2-1: regression still present, unchanged from 7.2.0-1.
Reverting to 7.1.8-1 (reinstalling the cached package) immediately restores full 4K@120Hz mode availability, confirming this is kernel-side and not a KWin/Plasma or TV/cable/EDID issue.


### Logs / stack traces

```text

Additional system information

Port the same approach used to fix #810 to DCN 3.2: replace the four REG_WAIT(DCHUBBUB_DETx_CTRL, ...) calls in dcn32_program_compbuf_size() (drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hubbub.c) with an equivalent dcn32_wait_for_det_apply()-style helper, mirroring dcn31_wait_for_det_apply(). Happy to test a patch on this hardware if one is provided.