From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: drm/vblank: timer: Fix timestamps and improve reliabilty Date: Thu, 04 Jun 2026 14:03:58 +1000 Message-ID: In-Reply-To: <20260601141922.91498-1-tzimmermann@suse.de> References: <20260601141922.91498-1-tzimmermann@suse.de> X-Mailer: Claude Code Patch Reviewer Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Overall Series Review Subject: drm/vblank: timer: Fix timestamps and improve reliabilty Author: Thomas Zimmermann Patches: 11 Reviewed: 2026-06-04T14:03:58.367915 --- This series fixes timestamps and improves reliability for the DRM vblank ti= mer infrastructure, used by virtual/software GPU drivers (vkms, bochs, virt= io, vmwgfx). The intent is sound =E2=80=94 the timestamp calculation has lo= ng-standing issues, and the improvements are worthwhile. However, **Patch 3 introduces a clock domain mismatch bug** that affects mu= ltiple subsequent patches. The timer is switched to `CLOCK_BOOTTIME`, but t= he initial timer start and various timestamp reads/comparisons still use `k= time_get()` / `ktime_get_ns()` which return `CLOCK_MONOTONIC` time. On any = system that has been suspended, boottime and monotonic time diverge, making= the comparisons invalid. This is the most critical issue in the series. Additionally, **Patch 1 does not update the vmwgfx caller** (`vmw_vkms_get_= vblank_timestamp`), which still ignores the new return value and will retur= n `true` with stale/uninitialized `vblank_time` on error. The refactoring patches (4, 7) are reasonable, and Patch 6 is a correct and= clear bugfix. Patch 2's timestamp correction logic is sound in principle. --- --- Generated by Claude Code Patch Reviewer