Claude review: drm/msm/dp: Add MST support for MSM chipsets

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

Overall Series Review

Subject: drm/msm/dp: Add MST support for MSM chipsets
Author: Yongxing Mou <yongxing.mou@oss.qualcomm.com>
Patches: 66
Reviewed: 2026-04-12T10:16:11.494271

---

This 39-patch series adds Multi-Stream Transport (MST) support to the Qualcomm MSM DP driver, enabling a single DisplayPort controller to drive up to 4 independent displays. The series is structured as: prerequisite fixes (patches 1-4), DP driver refactoring to support multi-stream abstractions (patches 5-28), sink-side MST initialization (patch 29), new MST panel/controller helpers (patches 30-32), and the core MST DRM bridge module (patches 33-39).

**Architecture**: The approach is sound — the series correctly uses `drm_dp_mst_topology_mgr` for topology management, `drm_private_obj` for per-bridge atomic state, and the standard part1/part2 payload allocation flow. The separation of link-level operations (`prepare`/`unprepare`) from stream-level operations (`enable`/`disable`) is well-designed for MST where multiple streams share a single trained link.

**Concerns**:

1. **Bug in patch 30**: `IS_ERR(dp->panel)` checks the wrong variable — should be `IS_ERR(dp_panel)`. This means a failed panel allocation will dereference an error pointer.

2. **State access issue in patch 34**: `atomic_check` uses `to_drm_dp_mst_topology_state(mst->mst_mgr.base.state)` which reads the *current committed* state rather than the new atomic state being validated. Should use `drm_atomic_get_mst_topology_state()` to get the state from the atomic transaction.

3. **Memory leak in patch 01**: `drm_edid_read_ddc()` result is never freed, and the old cleanup path (`drm_edid_free`) was removed.

4. **Hardcoded 100ms sleep in patch 29** for clearing MST state is excessive and not tied to any spec requirement.

5. **Missing error propagation**: Several functions (e.g. `msm_dp_mst_bridge_atomic_check`) compute `slots` from `drm_dp_atomic_find_time_slots()` but never check if `slots < 0` before returning success.

6. **Series ordering**: Some patches could be squashed (e.g., patches 13-14 both touch stream_id) and the series is long enough that reviewability suffers. The 39-patch count could be reduced.

Overall: The series needs at least the IS_ERR bug fix and the atomic_check state access fix before it can be merged. The architecture and general approach are correct.

---

---
Generated by Claude Code Patch Reviewer