Claude review: drm/amdgpu/userq: clean up VA state on create failure

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

Overall Series Review

Subject: drm/amdgpu/userq: clean up VA state on create failure
Author: Guangshuo Li <lgs201920130244@gmail.com>
Patches: 1
Reviewed: 2026-06-05T06:43:51.545614

---

**Verdict: The bug this patch tries to fix has already been fixed in drm-next through a larger refactor of `amdgpu_userq_create`.**

This is a single patch addressing a real bug in an older version of the code: when `amdgpu_userq_input_va_validate()` succeeds for one VA but a subsequent validation (or later create step) fails, the VA cursor linked onto `queue->userq_va_list` leaks and `bo_va->userq_va_mapped` remains stale. The analysis in the commit message is correct for the code it targets.

However, the patch is based on an older codebase that has since been substantially reworked on drm-next. The current tree (as of drm-next) shows:

1. `amdgpu_userq_create()` has already been restructured with proper unwind labels (`free_queue`, `free_fence_drv`, `clean_doorbell_bo`, `clean_mqd`, `erase_doorbell`).
2. The `userq_va_list` mechanism referenced by the patch no longer exists. VA tracking now uses a `userq_vas.va_array` union and the `userq_va_mapped` flag directly on `bo_va`.
3. The function `amdgpu_userq_buffer_vas_list_cleanup()` that the patch calls does not exist in the current tree.
4. The VA validation failure path already goes to `free_fence_drv` (line 662), which unwinds properly.
5. The short-circuit OR expression for VA validation now stores results into `&queue->userq_vas.va.*` output parameters rather than linking a list.

The patch cannot be applied to drm-next and is not needed there — the underlying issue has been addressed as part of the refactoring.

---
Generated by Claude Code Patch Reviewer