Claude review: arm64: dts: ti: beagley-ai: Enable HDMI display and audio

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

Board-level DTS enabling HDMI via IT66122 bridge on DSS1 VP1.

The DT looks correct: IT66122 at I2C address 0x4c, 24-bit DPI input, HDMI output to connector, audio via McASP1 in I2S mode.

**Minor:** The `dss1` node overrides `clocks` with:
```
clocks = <&k3_clks 232 8>, <&k3_clks 232 0>, <&k3_clks 232 4>;
```
This overrides the common dtsi's clocks which uses `<&dss1_vp1_clk>` for vp1. Since this board uses DPI (not OLDI), it provides the raw clock instead of the /7 divider. This is correct — the /7 divider is only needed for OLDI mode.

The `port@0` in the `dss1` ports is used for VP1 DPI output. The comment says "DSS1-VP1" which matches the hardware routing.

No issues.

---

**Summary of items requiring attention:**

1. **Patch 6 (critical):** The removal of `pm_runtime_mark_last_busy()` before `pm_runtime_put_autosuspend()` appears incorrect. Please verify that the autosuspend delay works as intended without it.
2. **Patch 12 (bug):** `WARN_ON(pm_runtime_get_sync())` will fire on success (returns 1). Use `pm_runtime_resume_and_get()` instead.
3. **Patch 12 (bug):** PM runtime reference leak on early-return error paths in `tidss_oldi_atomic_pre_enable()`.
4. **Patch 9 (design):** `syscon_regmap_lookup_by_compatible()` is fragile; consider phandle-based lookup. Error message is misleading.
5. **Patch 8 (minor):** `dpi_output` is written redundantly on every vp_setup call; could be set once.
6. **Patch 12 (minor):** Missing `ida_destroy()` at module exit.

---
Generated by Claude Code Patch Reviewer