Claude review: drm/amdgpu: fix sleeping allocation under spinlock in PASID IDR

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

Patch Review

**Status: Good**

The fix uses the standard `idr_preload()` / `GFP_NOWAIT` pattern:

```c
+	idr_preload(GFP_KERNEL);
 	spin_lock(&amdgpu_pasid_idr_lock);
 	pasid = idr_alloc_cyclic(&amdgpu_pasid_idr, NULL, 1,
-				 1U << bits, GFP_KERNEL);
+				 1U << bits, GFP_NOWAIT);
 	spin_unlock(&amdgpu_pasid_idr_lock);
+	idr_preload_end();
```

This is the textbook kernel approach: `idr_preload(GFP_KERNEL)` pre-allocates radix-tree nodes in sleepable context with preemption disabled, then `GFP_NOWAIT` inside the spinlock draws from the preloaded pool. `idr_preload_end()` re-enables preemption after the lock is released.

The ordering is correct: `idr_preload()` before `spin_lock()`, and `idr_preload_end()` after `spin_unlock()`. No issues.

Note: The Fixes: tag references `14b81abe7bdc` while the cover letter mentions `8f1de51f49be`. Both should refer to the same commit that introduced the IDR conversion. The cover letter and patch 2 use `8f1de51f49be` — the discrepancy in the Fixes: tag on patch 1 (`14b81abe7bdc`) should be verified. If these are different commits, the correct one to cite is whichever actually introduced `idr_alloc_cyclic()` with `GFP_KERNEL` under the spinlock.

---
Generated by Claude Code Patch Reviewer