Claude review: drm/bridge: it66121: Select HDMI or DVI mode based on sink type

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

Patch Review

**Status: One concern about state access in mode_set.**

The core logic is correct: checking `conn_state->connector->display_info.is_hdmi` in `atomic_check` and using that to conditionally skip AVI infoframes and write 0 (DVI mode) vs `IT66121_HDMI_MODE_HDMI` to the mode register.

The defensive defaulting is good:
```c
	/* Default to HDMI to preserve legacy behavior. */
	state->sink_is_hdmi = true;

	if (conn_state && conn_state->connector)
		state->sink_is_hdmi = conn_state->connector->display_info.is_hdmi;
```

**Concern: State access in `mode_set` via `bridge->base.state`.**

```c
	struct drm_bridge_state *bridge_state =
		drm_priv_to_bridge_state(bridge->base.state);
	struct it66121_bridge_state *state =
		to_it66121_bridge_state(bridge_state);
```

The `mode_set` callback has the old-style signature without a `drm_atomic_state *` parameter, so accessing the bridge state through `bridge->base.state` is the only option. This works because `mode_set` is called after the state swap during commit. However, this couples the driver to commit ordering assumptions.

**Suggestion:** Consider converting from the legacy `mode_set` callback to `atomic_enable` where the `struct drm_atomic_commit *state` parameter is available. The enable callback already exists — the mode_set logic could be folded into it, and you'd get the state cleanly via `drm_atomic_get_new_bridge_state()`. This would eliminate the fragile `bridge->base.state` access entirely. If that's too much churn for this fix, the current approach works but should have a brief comment noting why `bridge->base.state` is safe here (called post-swap during commit).

**Minor nit:** The AVI packet disable path writes 0 to `IT66121_AVI_INFO_PKT_REG`, which should explicitly clear the AVI info that was previously written. Depending on hardware behavior, you may also want to zero out the AVI infoframe data registers (`IT66121_AVIINFO_DB1_REG`) when switching to DVI mode, to avoid any stale data being accidentally transmitted if the packet enable bit is ever re-asserted without going through `mode_set` again. This is a hardening suggestion, not a correctness issue for the current flow.

---
Generated by Claude Code Patch Reviewer