Regression: reboot/shutdown hang (screen off, machine stays on) on linux-cachyos 7.1.5-1 and 7.2.rc5-1; 7.1.4-1 clean

### 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)
- linux-cachyos-rc

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

`cachyos-bugreport.sh` not run yet — system information provided manually below.

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

- Host: GMKtec, AMD Strix Halo (Ryzen AI Max 395, Radeon 8060S iGPU), 96 GiB
- OS: CachyOS (Arch-based, rolling), KDE Plasma 6, Wayland
- Displays: LG 32G810SA-WB (webOS, 4K) via USB-C DP Alt Mode (DP-3, active), primary LG via DP (DP-1, disconnected after reboots)
- FS: BTRFS (root + subvolumes incl. /var/cache)
- GRUB cmdline: `nowatchdog nvme_load=YES zswap.enabled=0 splash loglevel=3 amdgpu.exp_hw_support=1 iommu=pt amd_iommu=on amdgpu.noretry=0 pcie_aspm=off ttm.pages_limit=29884416 amdgpu.max_allocation_size=114688 nvme_core.default_ps_max_latency_us=0`

### Bug description

On `systemctl reboot` (also shutdown), system reaches "Reached target Shutdown", all services stop, journald stops — then the machine does NOT reboot: monitors go black (KDE/DRM turns them off first), fans stay running, USB stays powered. Looks like a hard hang. Requires forcing reboot via `systemctl reboot --force --force` or pressing power button. On the most recent hang (boot -1, 2026-08-01 00:53:14) journal shows "Reached target Shutdown", then nothing — kernel reboot never executes.

**Regression data (confirmed by bisecting installed kernels):**

- **linux-cachyos 7.1.4-1** (installed 2026-07-19): reboots clean, no hang. Multiple reboots confirmed 2026-08-02.
- **linux-cachyos 7.1.5-1** (installed 2026-07-27): hang. First confirmed hang 2026-07-31 (boot -7, ~5 min gap between boots).
- **linux-cachyos-rc 7.2.rc5-1** (installed 2026-08-01, stable 7.1.5-1 removed same day): hang on every reboot. Both reboots hung on "umount: /var/cache: target is busy" → "Failed unmounting /var/cache." → ~2.5 min stall. This was the known regression of issue #895.
- **linux-cachyos-lts 6.18.40-1**: not yet rebooted after fix; LTS unaffected by the 7.1 regression per issue #940/community.

So regression range: **between 7.1.4-1 and 7.1.5-1** (kernels released around 2026-07-27).

**What we already ruled out:**

- /var/cache mount busy (issue #895): we applied `LazyUnmount=yes` drop-in for var-cache.mount — /var/cache now unmounts cleanly. Hang still occurs AFTER unmount, at kernel reboot stage. Not the cause of the reboot hang (it WAS the cause of the earlier 2.5-min stall).
- tdp-tray / ryzen_smu DKMS module: code audit shows .remove only does kfree/kobject_put; process stops before shutdown target. Not the cause.
- ROCm 7.14 (user-space only): not the cause.
- reboot type: /sys/kernel/reboot/type = acpi (default). pstore empty (no crash log).

**Related issues on this repo:**

- #900 (CLOSED 2026-06-29): "Unable to fully shut down or restart since 7.1 kernel with AMD GPU and multiple displays" — SAME symptom (multiple displays; screen off, fans/USB still on). Community workarounds found there:
- USB power control fix: `for d in /sys/bus/pci/drivers/xhci_hcd/*:*; do echo on | sudo tee "$d/power/control"; done` (helped GSDragoon, KarenNatsumi)
- `amdgpu.dcdebugmask=0x800` kernel cmdline (helped 0ldMonk, 2026-07-28)
- Removing second display also avoids it.
- Note: our case uses USB-C DP Alt Mode to an LG webOS monitor — likely relevant (DP tunnel over USB-C / Thunderbolt-like path, and webOS sends hotplug/DP traffic; issue #900 referenced a linux-usb patch about USB-C/Thunderbolt, https://lore.kernel.org/linux-usb/20260611054157.928921-1-Basavaraj.Natikar@amd.com/).
- #903 (OPEN): "Motherboard fails to power off / hangs on shutdown (ASRock H410M-HDV)" — possibly related, unsolved.

**Why new issue:** #900 was closed as completed on 2026-06-29, but the regression returned/continued: on 7.1.5-1 and 7.2.rc5-1 (late July 2026) the hang reproduces on our system, while 7.1.4-1 is clean. This is a NEW regression window (7.1.4 → 7.1.5) worth tracking.

### Steps to reproduce

1. Boot linux-cachyos 7.1.5-1 or 7.2.rc5-1 (on hardware with AMD iGPU + USB-C DP monitor).
2. Run `systemctl reboot`.
3. Observe: services stop, "Reached target Shutdown" in journal, then hang (black screen, fans on, no reboot) — requires `--force --force` or power button.
4. On 7.1.4-1 same hardware: reboot completes normally.

### Expected behavior

`systemctl reboot` should complete the reboot. On 7.1.4-1 with the same hardware and configuration it does — the machine powers down/reboots cleanly.

### Actual behavior

Machine hangs at the kernel reboot stage after "Reached target Shutdown": monitors go black, fans keep spinning, USB stays powered, no reboot. Only `systemctl reboot --force --force` or the power button recovers it.

### Logs / stack traces

Journal up to the hang (boot -1, 2026-08-01 00:53:14):

```
Reached target Shutdown.

```

pstore is empty (no crash log). Kernel cmdline includes `loglevel=3`; let me know if a more verbose journal (`loglevel=7`/`initcall_debug`) or `cachyos-bugreport.sh` output is needed.

### Additional system information

- RAM: 96 GiB, desktop: KDE Plasma 6 (Wayland), init: systemd, bootloader: GRUB.
- The monitor is connected via USB-C DP Alt Mode (USB4/Thunderbolt-like path) — hotplug/DP traffic from the LG webOS monitor may be relevant (see #900 references).
- Reboot type: `/sys/kernel/reboot/type` = `acpi` (default).

### Request

- Please identify the commit between 7.1.4-1 and 7.1.5-1 (amdgpu/DRM/USB) that reintroduced the shutdown hang.
- Please advise whether the community workarounds (USB power control, `amdgpu.dcdebugmask=0x800`) address this regression.

Start by reproducing the shutdown hang on linux-cachyos 7.1.5-1 or 7.2.rc5-1 and compare it with 7.1.4-1, focusing on the amdgpu/DRM and USB-C display paths described here and in issue #900. Check the regression range for the responsible change, then test whether the USB power-control workaround or amdgpu.dcdebugmask=0x800 avoids the hang; done means identifying the change or confirming a workaround.

Written by the indexing model from the issue text.