Claude review: rust: drm: Pass bound parent device to ioctl handlers

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

Patch Review

The final ioctl call becomes:
```rust
match $func(dev, &*guard, guard.registration_data(), data, file) {
```

**`&*guard` usage**: This dereferences the `UnbindGuard` through its `Deref` impl to get `&T::ParentDevice<Bound>`, then re-borrows it. This is correct.

**Borrow concern**: Both `&*guard` and `guard.registration_data()` borrow the guard simultaneously — the first immutably through `Deref`, the second through a method call. Since both are immutable borrows, this is fine from a borrow-checking perspective. However, `registration_data()` internally calls `self.dev.raw_registration_data()`, which goes through `UnsafeCell::get()`. The safety here relies on the fact that no mutable access to `registration_data` occurs during the SRCU critical section, which is guaranteed because `Registration::drop` calls `drm_dev_unplug()` (SRCU barrier) before modifying anything.

**Driver-side changes**: Nova and tyr add the parent device parameter (`_adev: &auxiliary::Device<Bound>` and `_pdev: &platform::Device<device::Bound>` respectively). These are unused with `_` prefix, which is expected for this infrastructure patch.

**Argument order**: The final ioctl handler signature is:
```
fn foo(dev, parent_device, reg_data, ioctl_data, file) -> Result<u32>
```
This is a 5-argument function. While not a problem per se, it's worth considering if a context struct would be cleaner for future evolution. But for now, this is fine.

**Minor nit**: The imports in `nova/file.rs` add `auxiliary` and `device::Bound` — straightforward.

No blocking issues.

### Summary

The series is well-designed with careful safety arguments. The main architectural dependency is on the `ForLt` trait and `DeviceContext` work that isn't in drm-next yet. The `drm_dev_unregister` → `drm_dev_unplug` switch is the key safety enabler and is correct. The lifetime erasure/recovery pattern via `mem::transmute` is the most subtle part but is justified by `ForLt`'s covariance guarantee. The incremental ioctl handler signature evolution across patches 4-6 is clean. No blocking issues found.

---
Generated by Claude Code Patch Reviewer