Claude review: drm/vblank: timer: Fix timestamp calculation

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

Patch Review

**Concept:** Correct. The returned timestamp should be the start of the active display period, not the start of the vblank period. Subtracting `activedur_ns` instead of `framedur_ns` (i.e., `vtimer->interval`) achieves this.

**The math:**
```c
activedur_ns = div_s64(framedur_ns * mode->crtc_vdisplay, mode->crtc_vtotal);
*vblank_time = ktime_sub_ns(*vblank_time, activedur_ns);
```

`next_expiry - active_duration = start_of_active_period`. This is correct geometry.

**Observation on `drm_WARN_ON(dev, !mode->crtc_vtotal)`:** This is a good safety check, but after this returns `false`, the caller (`drm_crtc_vblank_helper_get_vblank_timestamp_from_timer`) now also returns `false`. Make sure callers handle that gracefully — this may trigger DRM's "failed to get vblank timestamp" warnings. A `vtotal` of 0 indicates a fundamentally broken mode, so warning is appropriate.

**No issues with the logic itself.** This is a real bug fix for a long-standing issue.

---

---
Generated by Claude Code Patch Reviewer