Claude review: drm/mode-config: Create drm_mode_config_create_initial_state()

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

Patch Review

**Bug: connector list iterator leaked on error.**

In `drm_mode_config_create_initial_state()`:

```c
drm_connector_list_iter_begin(dev, &conn_iter);
drm_for_each_connector_iter(connector, &conn_iter) {
    if (connector->state)
        continue;

    ret = drm_mode_config_connector_create_state(connector);
    if (ret)
        return ret;  // <-- BUG: misses drm_connector_list_iter_end()
}
drm_connector_list_iter_end(&conn_iter);
```

If `drm_mode_config_connector_create_state()` returns an error, the function returns without calling `drm_connector_list_iter_end(&conn_iter)`. The iterator holds a reference that won't be released. This needs to be fixed, e.g.:

```c
    if (ret) {
        drm_connector_list_iter_end(&conn_iter);
        return ret;
    }
```

Additionally, there's **no cleanup of already-allocated states on error**. If, say, plane states are created successfully but a CRTC state allocation fails, the already-created plane states remain allocated and dangling. Whether this matters depends on whether the caller tears down the device on failure — the skeleton in patch 17 shows `return ret` which would trigger `drmm_mode_config_init` cleanup via devres, which should handle it. But it's worth considering explicitly.

Also, the function skips colorop `atomic_create_state` — it calls `drm_atomic_helper_colorop_create_state()` directly, which means drivers cannot override colorop state creation. This is consistent since colorops don't have an `atomic_create_state` callback in `drm_colorop_funcs` (the callback added in patch 7 is a free function, not a vtable entry). Just noting the asymmetry.

---
Generated by Claude Code Patch Reviewer