Claude review: drm: writeback: Refactor drm_writeback_connector structure

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

Patch Review

This is the core patch. The structural change is sound: `drm_writeback_connector` loses its `struct drm_connector base` member and instead gets embedded inside `struct drm_connector` in a union with `hdmi`. The helper `drm_connector_to_writeback()` is replaced with the inverse `drm_writeback_to_connector()` using `container_of`.

**Issues:**

1. **Typo in doc comment** — the `@writeback` kdoc says "valriables" instead of "variables":
   ```c
   /**
    * @writeback: Writeback related valriables.
    */
   struct drm_writeback_connector writeback;
   ```

2. **`drm_writeback_to_connector()` is only safe if the writeback is actually embedded in a `drm_connector`** — the `container_of` relies on the `writeback` field being part of `drm_connector`. This is now always true by construction, but there's no runtime assertion. The existing `WARN_ON` in `drm_writeback_get_out_fence` checking `connector_type` is good defensive practice; consider whether `drm_writeback_to_connector()` itself should have a similar check.

3. **Inconsistency in `dm_set_writeback`** — the amdgpu code changes `drm_connector_to_writeback(connector)` to a direct `&connector->writeback` access:
   ```c
   -	struct drm_writeback_connector *wb_conn = drm_connector_to_writeback(connector);
   +	struct drm_writeback_connector *wb_conn = &connector->writeback;
   ```
   This is fine functionally, but inconsistent with other places in the same patch that use the new `drm_writeback_to_connector()` helper. Since the helper now goes the other direction (wb -> connector), this is correct but could use `&connector->writeback` consistently throughout.

4. **malidp_hw.c changes** — This file is modified in patch 1 (changing `drm_writeback_signal_completion` calls from `&malidp->mw_connector` to `&malidp->mw_connector.writeback`) and then again in patch 4 (changing them back from `.writeback` to taking the connector directly). This creates a transient churn. Consider squashing or reordering to avoid changing these lines twice.

5. **vkms_writeback.c `vkms_wb_cleanup_job`** — Uses a raw `container_of` instead of `drm_writeback_to_connector()`:
   ```c
   struct drm_connector *connector = container_of(wb_conn,
                                                   struct drm_connector,
                                                   writeback);
   ```
   This is functionally identical to `drm_writeback_to_connector()` but inconsistent. Since patch 7 removes this entirely, it's a transient issue but still worth noting for bisectability.

---
Generated by Claude Code Patch Reviewer