Claude review: drm/vblank: Add DRM_VBLANK_FLAG_SIMULATED

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

Patch Review

This is the core behavioral patch. It defines `DRM_VBLANK_FLAG_SIMULATED` as `BIT(1)` and adds early-return checks in three ioctl handlers.

**Observations:**

- **BIT(1) not BIT(0):** The flag starts at `BIT(1)` rather than `BIT(0)`. This is unusual but perhaps intentional to reserve BIT(0) for something else, or a minor mistake. It works correctly either way since the flags field starts at 0, but it's worth asking whether this was deliberate. If there's no planned BIT(0), starting at BIT(0) would be more conventional.

- The three ioctl checks are placed at sensible positions — after the vblank struct is obtained but before any real work is done:

  In `drm_wait_vblank_ioctl`:
  ```c
  +	if (vblank->flags & DRM_VBLANK_FLAG_SIMULATED)
  +		return -EOPNOTSUPP;
  ```

  In `drm_crtc_get_sequence_ioctl`:
  ```c
  +	if (vblank->flags & DRM_VBLANK_FLAG_SIMULATED)
  +		return -EOPNOTSUPP;
  ```

  In `drm_crtc_queue_sequence_ioctl`:
  ```c
  +	if (vblank->flags & DRM_VBLANK_FLAG_SIMULATED)
  +		return -EOPNOTSUPP;
  ```

- **Error code choice:** `-EOPNOTSUPP` is reasonable for "this operation is not supported on this CRTC." The alternative would be `-ENOTSUP` (same value on Linux) or `-EINVAL`. The choice here is fine — it clearly communicates that the operation itself is understood but not available for this device.

- **Potential concern in `drm_crtc_queue_sequence_ioctl`:** The flag check is placed *after* the `kzalloc_obj(*e)` call... wait, re-reading the diff more carefully:
  ```c
  +	if (vblank->flags & DRM_VBLANK_FLAG_SIMULATED)
  +		return -EOPNOTSUPP;
  +
   	e = kzalloc_obj(*e);
  ```
  Actually the check is placed *before* the allocation. Good — this avoids allocating and then immediately freeing.

- **UAPI impact:** This changes behavior for existing userspace applications. Programs that currently call `DRM_IOCTL_WAIT_VBLANK` on these virtual GPU devices will start getting `-EOPNOTSUPP` instead of (potentially misleading) success. The cover letter notes that major compositors (Weston, KWin, Mutter) don't rely on this ioctl for rate-limiting, and this was discussed on IRC. This seems adequately vetted.

---
Generated by Claude Code Patch Reviewer