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: rust: gpuvm: add GpuVm::obtain()
Date: Sun, 22 Mar 2026 03:35:23 +1000
Message-ID: <review-patch3-20260320-gpuvm-rust-v5-3-76fd44f17a87@google.com>
In-Reply-To: <20260320-gpuvm-rust-v5-3-76fd44f17a87@google.com>
References: <20260320-gpuvm-rust-v5-0-76fd44f17a87@google.com> <20260320-gpuvm-rust-v5-3-76fd44f17a87@google.com>
X-Mailer: Claude Code Patch Reviewer
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

Patch Review

Introduces `GpuVmBo<T>`, `GpuVmBoAlloc<T>`, and the `obtain()` path.

**Key concern =E2=80=94 resv lock in `obtain()`:**
```rust
if ptr::eq(ptr, me.as_raw()) && me.gpuvm().is_extobj(me.obj()) {
    let resv_lock =3D me.gpuvm().raw_resv();
    // TODO: Use a proper lock guard here once a dma_resv lock abstraction =
exists.
    unsafe { bindings::dma_resv_lock(resv_lock, ptr::null_mut()) };
    unsafe { bindings::drm_gpuvm_bo_extobj_add(ptr) };
    unsafe { bindings::dma_resv_unlock(resv_lock) };
}
```
Passing `null` for `ww_acquire_ctx` means this cannot participate in deadlo=
ck avoidance. If a driver calls `obtain()` while already holding another re=
sv lock in a ww transaction, this will deadlock. The TODO acknowledges this=
 but it should be documented more prominently as a limitation, possibly wit=
h a `// WARNING` or a doc comment on `obtain()` itself.

**`GpuVmBoAlloc::Drop` =E2=80=94 could use `drm_gpuvm_bo_destroy_not_in_lis=
ts()`:**
```rust
fn drop(&mut self) {
    // TODO: Call drm_gpuvm_bo_destroy_not_in_lists() directly.
    unsafe { bindings::drm_gpuvm_bo_put_deferred(self.as_raw()) };
}
```
The TODO is noted. Using `drm_gpuvm_bo_put_deferred` here works but does un=
necessary refcount work for an object known to have refcount=3D1 and not be=
 in any lists.

**ALLOC_FN optimization is clever:**
```rust
pub(super) const ALLOC_FN: Option<unsafe extern "C" fn() -> *mut bindings::=
drm_gpuvm_bo> =3D {
    ...
    if base.size() !=3D rust.size() || base.align() !=3D rust.align() {
        Some(Self::vm_bo_alloc)
    } else {
        None
    }
};
```
This correctly avoids overriding when the Rust wrapper has no extra fields.=
 However, the `FREE_FN` check uses `needs_drop::<Self>()` which is independ=
ent =E2=80=94 when `VmBoData` has a trivial type but non-default layout, yo=
u'd get `ALLOC_FN =3D Some` but `FREE_FN =3D None`, meaning the C side woul=
d `kfree()` a larger allocation which is fine since it was `kmalloc`'d. Thi=
s works correctly.

---
Generated by Claude Code Patch Reviewer