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: gpu: nova-core: send UNLOADING_GUEST_DRIVER GSP command upon unloading
Date: Thu, 23 Apr 2026 08:52:55 +1000
Message-ID: <review-patch4-20260421-nova-unload-v2-4-2fe54963af8b@nvidia.com>
In-Reply-To: <20260421-nova-unload-v2-4-2fe54963af8b@nvidia.com>
References: <20260421-nova-unload-v2-0-2fe54963af8b@nvidia.com> <20260421-nova-unload-v2-4-2fe54963af8b@nvidia.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

This is the core patch. It adds:

1. **Bindings** (`gsp/fw/r570_144/bindings.rs`): Adds `rpc_unloading_guest_=
driver_v1F_07` struct.

2. **Firmware command wrapper** (`gsp/fw/commands.rs`):
```rust
+pub(crate) struct UnloadingGuestDriver(bindings::rpc_unloading_guest_drive=
r_v1F_07);
+
+impl UnloadingGuestDriver {
+    pub(crate) fn new(suspend: bool) -> Self {
+        Self(bindings::rpc_unloading_guest_driver_v1F_07 {
+            bInPMTransition: u8::from(suspend),
+            bGc6Entering: 0,
+            newLevel: if suspend { 3 } else { 0 },
+            ..Zeroable::zeroed()
+        })
+    }
+}
```

The magic number `3` for `newLevel` when suspending appears to correspond t=
o a D3 power state. This matches what OpenRM does (`NV2080_CTRL_GPU_SET_POW=
ER_STATE_GPU_LEVEL_3`). A named constant would be nice but not critical sin=
ce this is matching hardware/firmware protocol values.

3. **GSP command** (`gsp/commands.rs`): Implements `CommandToGsp` and `Mess=
ageFromGsp` for the new command. The reply type `UnloadingGuestDriverReply`=
 has `type Message =3D ()` =E2=80=94 the reply payload is empty/ignored, wh=
ich is correct for this RPC.

4. **Shutdown logic** (`gsp/boot.rs`):
```rust
+    fn shutdown_gsp(
+        cmdq: &Cmdq,
+        bar: &Bar0,
+        gsp_falcon: &Falcon<Gsp>,
+        suspend: bool,
+    ) -> Result<()> {
+        cmdq.send_command(bar, commands::UnloadingGuestDriver::new(suspend=
))?;
+
+        const LIBOS_INTERRUPT_PROCESSOR_SUSPENDED: u32 =3D 0x8000_0000;
+        read_poll_timeout(
+            || Ok(gsp_falcon.read_mailbox0(bar)),
+            |&mb0| mb0 =3D=3D LIBOS_INTERRUPT_PROCESSOR_SUSPENDED,
+            Delta::from_millis(10),
+            Delta::from_secs(5),
+        )
+        .map(|_| ())
+    }
```

The 5-second timeout with 10ms polling interval seems reasonable. The `LIBO=
S_INTERRUPT_PROCESSOR_SUSPENDED` constant matches what OpenRM uses (`NV_LIB=
OS_INTERRUPT_PROCESSOR_SUSPENDED`).

5. **Integration** (`gpu.rs`):
```rust
+        let _ =3D kernel::warn_on_err!(self.gsp.unload(dev, bar, &self.gsp=
_falcon));
```

The `let _ =3D` discard after `warn_on_err!` means if unload fails, we cont=
inue with `sysmem_flush.unregister()`. This is the right approach for an un=
bind path =E2=80=94 best-effort teardown.

**Minor observation:** `shutdown_gsp` is a static method (`&Cmdq` parameter=
 instead of `&self`) while `unload` takes `&self`. This is fine since `shut=
down_gsp` is a private helper, but worth noting that `cmdq` comes from `sel=
f.cmdq` in the caller.

No blocking issues.

---
Generated by Claude Code Patch Reviewer