Claude review: drm/gem: Try to fix change_handle ioctl, attempt 4

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

Overall Series Review

Subject: drm/gem: Try to fix change_handle ioctl, attempt 4
Author: Simona Vetter <simona.vetter@ffwll.ch>
Patches: 3
Reviewed: 2026-06-05T06:00:21.529702

---

This mbox contains two versions (v1 and v2) of the same single patch. The v2 supersedes v1, with the key change being the use of `IS_ERR_OR_NULL()` instead of `!obj` for the `idr_replace()` return value check — a critical fix since `idr_replace()` returns `ERR_PTR(-ENOENT)` for non-existent IDs, not NULL.

The patch addresses a genuinely tangled mess: three stacked bugfix commits (`5e28b7b94408`, `7164d78559b0`, `dc366607c41c`) each partially wrong due to confusing variable naming (`handle` actually meant `new_handle`). The approach is sound — it rebuilds the function using the two established patterns from the same file:

1. **`drm_gem_handle_delete`'s pattern**: `idr_replace(NULL, handle)` to atomically grab the object and NULL out the old slot
2. **`drm_gem_handle_create_tail`'s pattern**: `idr_alloc(NULL, ...)` then later `idr_replace(obj, ...)` for the two-stage allocation

Disabling the ioctl via `drm_invalid_op` until IGT tests and consensus are in place is a reasonable defensive measure given the history of repeated incorrect fixes.

The v2 patch is **correct** and significantly cleaner than the existing code. I have only minor observations below.

---
Generated by Claude Code Patch Reviewer