Claude review: drm/tidss: Add BeagleY-AI display support (and some more)

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

Overall Series Review

Subject: drm/tidss: Add BeagleY-AI display support (and some more)
Author: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Patches: 16
Reviewed: 2026-04-23T09:45:56.852621

---

This is a well-structured 15-patch series that adds display support for TI AM62P/J722S/AM67A SoCs and the BeagleY-AI board. The series covers DT bindings, driver enhancements (DPI detection, signal edge config, DPIENABLE, OLDI polarity fixes), a significant OLDI-to-auxiliary-device conversion, AM62P device support, and board-level DTS for BeagleY-AI HDMI output.

The overall architecture is sound. The OLDI auxiliary bus conversion (patch 12) is the most complex change and is generally well done — it cleanly separates OLDI into a proper device model enabling per-OLDI power domain management. The DPI detection mechanism (patch 8) is pragmatic given the hardware constraints.

**Key concerns:**

1. **`WARN_ON(pm_runtime_get_sync())` in patch 12** — `pm_runtime_get_sync()` returns a positive value (1) on success when the device was already active. `WARN_ON()` treats any non-zero as true, so this will fire a spurious warning on every pre_enable after the first. Should use `pm_runtime_resume_and_get()` or check `< 0`.

2. **`syscon_regmap_lookup_by_compatible()` in patch 9** — This is a global lookup by compatible string rather than by phandle. On SoCs with two DSS instances (AM62P), both DSS instances would find the same syscon node. This happens to be correct if the register is shared, but it's fragile and not the usual pattern. The DT binding (patch 5) doesn't add a `ti,dpi-io-ctrl` phandle property to the DSS binding, though the DTS (patch 14) does reference it. Consider using `syscon_regmap_lookup_by_phandle()` for robustness.

3. **`dpi_output` stored redundantly** — The `dpi_output` flag is set per-call in `dispc_vp_setup()` via a parameter from `tcrtc->dpi_output`, stored into `dispc->vp_data[hw_videoport].dpi_output`. This value never changes after CRTC creation, so writing it on every `dispc_vp_setup()` call is unnecessary; it should be set once during init.

4. **Missing `of_node_put` for `oldi_tx` on success path in `for_each_available_child_of_node` loop** in patch 12's `tidss_oldi_create_devices()` — the loop's `oldi_tx` reference is put on error, but `of_node_put(oldi_tx)` is not called explicitly on a normal break/return. The `for_each_available_child_of_node` macro only puts the node automatically when the loop runs to completion, which it does here (falling through), so this is actually fine.

---

---
Generated by Claude Code Patch Reviewer