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: drm/amdgpu: amdgpu{_reset}.h: fix all kernel-doc warnings
Date: Thu, 05 Mar 2026 13:13:38 +1000
Message-ID: <review-patch1-20260305005556.1222863-1-rdunlap@infradead.org>
In-Reply-To: <20260305005556.1222863-1-rdunlap@infradead.org>
References: <20260305005556.1222863-1-rdunlap@infradead.org> <20260305005556.1222863-1-rdunlap@infradead.org>
X-Mailer: Claude Code Patch Reviewer
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

Patch Review

**Doc fix for `amdgpu_wb` =E2=80=94 Correct:**

```c
- * amdgpu_wb - This struct is used for small GPU memory allocation.
+ * struct amdgpu_wb - This struct is used for small GPU memory allocation.
```

Straightforward and correct. The `struct` keyword is required by kernel-doc.

**Enum doc name fixes =E2=80=94 Correct:**

The doc comments now use `@AMD_RESET_METHOD_LEGACY`, `@AMD_RESET_METHOD_MOD=
E0`, etc., matching the actual enum member names. Previously the docs used =
shortened names (`@AMD_RESET_LEGACY`, `@AMD_RESET_MODE0`) that didn't match=
. The new `@AMD_RESET_METHOD_ON_INIT` description is also a good addition =
=E2=80=94 it was previously undocumented.

**Circular include / forward reference issue =E2=80=94 Problem:**

`amdgpu_reset.h` includes `amdgpu.h` at line 27:
```c
#include "amdgpu.h"
```

The patch adds `#include "amdgpu_reset.h"` near line 1335 of `amdgpu.h`:
```c
+#include "amdgpu_reset.h"
```

But `enum amd_reset_method` is used at `amdgpu.h:572`:
```c
enum amd_reset_method (*reset_method)(struct amdgpu_device *adev);
```

When a compilation unit includes `amdgpu.h` first, the compiler reaches lin=
e 572 (the `reset_method` function pointer using `enum amd_reset_method`) b=
efore the `#include "amdgpu_reset.h"` at line 1335 where the enum is now de=
fined. The enum type is used ~760 lines before it becomes visible.

GCC may tolerate this as a GNU extension (implicit forward declaration of t=
he enum), but it's not clean C and may produce warnings with stricter compi=
lers. A better approach would be to either:
- Place the `#include "amdgpu_reset.h"` before the struct that uses the enu=
m (but this requires care with the circular dependency), or
- Add a forward declaration of the enum before its first use, or
- Break the circular dependency by having `amdgpu_reset.h` not include `amd=
gpu.h` and instead use forward declarations for the types it needs.

**Unexplained struct reorder in `amdgpu_reset.h`:**

The patch moves `struct amdgpu_reset_control` above `struct amdgpu_reset_ha=
ndler`. This reordering is not mentioned in the commit message and doesn't =
fix any kernel-doc warning. Both structs reference each other only through =
pointers, so the order doesn't matter for compilation, but the change adds =
unnecessary noise to the diff. If there's a reason for this move, it should=
 be documented in the commit message.

**Minor nit in comment text:**

The MODE0 description still has "the any device" (carried over from the ori=
ginal):
```
 * @AMD_RESET_METHOD_MODE0: Reset the entire ASIC. Not currently available =
for
 *                          the any device.
```
Should be "any device" (drop the extra "the"). Since this patch is already =
touching these comments, it would be a good opportunity to fix this typo.

---
Generated by Claude Code Patch Reviewer