Claude review: drm/atomic: Rework initial state allocation

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

Overall Series Review

Subject: drm/atomic: Rework initial state allocation
Author: Maxime Ripard <mripard@kernel.org>
Patches: 20
Reviewed: 2026-05-27T14:30:48.205573

---

This is a well-structured 19-patch series from Maxime Ripard that introduces a new `atomic_create_state` callback pattern for DRM planes, CRTCs, and connectors, mirroring what already exists for `drm_private_obj`. The motivation is sound: `reset()` is overloaded (initial state allocation, hardware reset, suspend/resume recovery), non-fallible, and behaves differently for different object types. Separating pristine state allocation into `atomic_create_state` makes the initialization path cleaner and fallible.

The series is on its 6th revision and has accumulated Reviewed-by tags from Thomas Zimmermann, Laurent Pinchart, Dmitry Baryshkov, and Tomi Valkeinen across most patches. The incremental approach (doc fixes, renames, then new callbacks, then helpers, then driver conversions) is methodical and makes bisection clean.

**Key concerns:**

1. **Missing cleanup on partial failure in `drm_mode_config_create_initial_state()`** (patch 16): If allocation fails partway through (e.g., a connector state allocation fails after planes and CRTCs succeeded), the already-allocated states are leaked. There's no rollback path.

2. **Minor doc typo**: Patch 12 has a double space in "Allocates and  initializes" in the `drm_atomic_helper_crtc_create_state()` doc. Also present in patch 15's `drm_atomic_helper_connector_create_state()`.

3. **`drm_atomic_commit` vs `drm_atomic_state`**: The doc in patch 1 references `drm_atomic_commit` and `drm_atomic_commit_alloc()`. The upstream tree currently uses `drm_atomic_state` — this appears to be part of a separate renaming that must be in the same tree. This is fine if the base commit includes that rename, but worth noting for readers.

Overall, this is a clean and well-reasoned series. The main substantive issue is the partial-failure leak in `drm_mode_config_create_initial_state()`.

---

---
Generated by Claude Code Patch Reviewer