[dmesg-broken-7.2.5-3-omarchy.log](https://github.com/user-attachments/files/32594604/dmesg-broken-7.2.5-3-omarchy.log)
[dmesg-working-7.2.3-arch1-3.log](https://github.com/user-attachments/files/32594602/dmesg-working-7.2.3-arch1-3.log)

# amdgpu MST link enable failure on kernel 7.2.5 (regression from 7.2.3) — external monitor stays black via USB-C dock

## Summary

After upgrading from kernel 7.2.3 to 7.2.5, external monitors connected via a
USB-C dock (DisplayPort MST, 2-lane link) fail to enable. The monitor is
detected and EDID is read correctly, but every modeset fails with
`enabling link 2 failed: 15` and a kernel WARNING in
`fill_dc_mst_payload_table_from_drm`. The connector stays at `enabled=disabled`,
`dpms=Off`, mode `0x0`. Rolling back to 7.2.3 resolves the issue.

## Hardware

- **Laptop:** Lenovo ThinkPad X1 Carbon (21ACS0L700), BIOS R1QET47W (1.33)
- **GPU:** AMD Cezanne — Radeon Vega Mobile (1002:1638), DCN 2.1, amdgpu driver
- **Dock:** Lenovo ThinkPad USB-C Dock Gen2 (uses DisplayPort MST, 2-lane link)
- **Monitor:** GIGA-BYTE G34WQC, 3440x1440 ultrawide, connected via DP through the dock

## Kernel versions

- **Broken:** `7.2.5-3-omarchy` (linux-omarchy 7.2.5-3, based on Linux 7.2.5)
- **Working:** `7.2.3-arch1-3` (Arch linux 7.2.3, based on Linux 7.2.3)

## Steps to reproduce

1. Boot kernel 7.2.5 with the USB-C dock connected and an external monitor attached via DP
2. Monitor is detected (EDID read, modes enumerated) but stays black
3. `hyprctl monitors` shows the external monitor at `0x0` (no mode applied)
4. Kernel connector `card1-DP-4` shows `status=connected`, `enabled=disabled`, `dpms=Off`

## Expected behavior

External monitor enables at 3440x1440@60 (or any mode), as it does on kernel 7.2.3.

## Actual behavior on 7.2.5

Connector detected but never enabled. Kernel log shows repeated failures:

```
[drm] DM_MST: Differing MST start on aconnector: 00000000d58618fe [id: 135]
[drm] Downstream port present 1, type 2
[drm] pre_validate_dsc:1746 MST_DSC dsc precompute is not needed
amdgpu 0000:07:00.0: [drm] enabling link 2 failed: 15
[drm] pre_validate_dsc:1821 MST_DSC crtc[1] needs mode_change
WARNING: drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_helpers.c:275 at fill_dc_mst_payload_table_from_drm+0x191/0x1a0 [amdgpu]
```

The `enabling link 2 failed: 15` repeats on every attempt. The link is
negotiated as `DP14, 2-lane` (the dock splits lanes for USB + display).

## Comparison: kernel 7.2.3 (working)

Same hardware, same dock, same monitor. The MST path is exercised
identically but does NOT fail:

```
[drm] DM_MST: Differing MST start on aconnector: 0000000043b93038 [id: 135]
[drm] Downstream port present 1, type 2
[drm] pre_validate_dsc:1746 MST_DSC dsc precompute is not needed
[drm] pre_validate_dsc:1821 MST_DSC crtc[1] needs mode_change
```

No `enabling link 2 failed: 15`, no `fill_dc_mst_payload_table_from_drm`
WARNING. Monitor enables normally.

## Kernel WARNING / call trace (7.2.5)

```
WARNING: drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm_helpers.c:275 at fill_dc_mst_payload_table_from_drm+0x191/0x1a0 [amdgpu], CPU#0: plymouthd/249
CPU: 0 UID: 0 PID: 249 Comm: plymouthd Not tainted 7.2.5-3-omarchy #1 PREEMPT(full)
Hardware name: LENOVO 21ACS0L700/21ACS0L700, BIOS R1QET47W (1.33 ) 11/15/2023
RIP: 0010:fill_dc_mst_payload_table_from_drm+0x191/0x1a0 [amdgpu]

Call Trace:
 <TASK>
 dm_helpers_dp_mst_write_payload_allocation_table+0x1a0/0x220 [amdgpu]
 link_set_dpms_off+0x674/0x710 [amdgpu]
 dcn20_reset_back_end_for_pipe+0xc6/0x370 [amdgpu]
 dcn20_reset_hw_ctx_wrap+0x112/0x190 [amdgpu]
 dce110_apply_ctx_to_hw+0x70/0x390 [amdgpu]
 dc_commit_state_no_check+0x5de/0xf20 [amdgpu]
 dc_commit_streams+0x297/0x560 [amdgpu]
 amdgpu_dm_atomic_commit_tail+0x33d/0x3680 [amdgpu]
 drm_atomic_helper_commit+0x160/0x1b0
 drm_atomic_commit+0xb1/0xf0
 drm_atomic_helper_set_config+0x96/0x110
 drm_mode_setcrtc+0x53e/0x9f0
 drm_ioctl_kernel+0xb6/0x110
 drm_ioctl+0x2e9/0x550
 amdgpu_drm_ioctl+0x4a/0x80 [amdgpu]
 <TASK>
```

## Compositor / environment

- Wayland, Hyprland 0.56.2 (not relevant — the failure is at the kernel modeset level; Hyprland reports the monitor correctly but the kernel rejects the modeset)
- amdgpu module params: dc=-1 (auto), dcdebugmask=0 (DSC enabled), modeset=-1

## Attachments

Full dmesg from both kernels is available:
- `dmesg-broken-7.2.5-3-omarchy.log` — full kernel log from the broken boot
- `dmesg-working-7.2.3-arch1-3.log` — full kernel log from the working boot