From mboxrd@z Thu Jan  1 00:00:00 1970
From: Claude Code Review Bot <claude-review@example.com>
To: dri-devel-reviews@example.com
Subject: Claude review: accel/neutron: Add GEM buffer object support
Date: Mon, 09 Mar 2026 08:49:22 +1000
Message-ID: <review-patch5-20260306-neutron-v2-5-3019bd8c91ef@nxp.com>
In-Reply-To: <20260306-neutron-v2-5-3019bd8c91ef@nxp.com>
References: <20260306-neutron-v2-0-3019bd8c91ef@nxp.com> <20260306-neutron-v2-5-3019bd8c91ef@nxp.com>
X-Mailer: Claude Code Patch Reviewer
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0

Patch Review

**SYNC_BO IOCTL direction semantics:** The IOCTL uses `DRM_IOWR` but `NEUTRON_SYNC_BO` doesn't write anything back to userspace. It should be `DRM_IOW`:
```c
#define DRM_IOCTL_NEUTRON_SYNC_BO \
	DRM_IOWR(DRM_COMMAND_BASE + DRM_NEUTRON_SYNC_BO, \
		 struct drm_neutron_sync_bo)
```

**Missing `pad` fields validation in `drm_neutron_inference_job`:** The struct has `__u32 pad[5]` but there's no check that these are zero (MBZ), yet they're documented as MBZ.

**EFAULT vs ENOMEM for alignment check:**
```c
if (drm_WARN_ON(drm, !IS_ALIGNED(dma_obj->dma_addr, NEUTRON_BO_ALIGN))) {
	ret = -EFAULT;
```
`-EFAULT` means bad address in user context. `-ENOMEM` or `-EIO` would be more appropriate for a kernel-side alignment failure.

---
Generated by Claude Code Patch Reviewer