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/vkms: Backport generic vblank timer to fix ABBA deadlock Date: Tue, 26 May 2026 07:07:30 +1000 Message-ID: In-Reply-To: <20260525131610.608273-1-w15303746062@163.com> References: <20260525131610.608273-1-w15303746062@163.com> 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/vkms: Backport generic vblank timer to fix ABBA deadlock Author: w15303746062@163.com Patches: 12 Reviewed: 2026-05-26T07:07:30.521650 --- This mbox contains two submissions from Mingyu Wang: 1. **A standalone patch** (May 15) =E2=80=94 a minimal fix using `hrtimer_t= ry_to_cancel()` to address an ABBA deadlock in the vkms vblank timer. 2. **A 5-patch series** (May 25) =E2=80=94 cherry-picks of Thomas Zimmerman= n's mainline commits that backport the generic DRM vblank timer infrastruct= ure to 6.18.y, replacing the custom vkms timer entirely. The series is the correct approach, as confirmed by discussion with Greg KH= and Maarten Lankhorst on the mailing list. The standalone patch was the ea= rlier attempt that was superseded. The cherry-picks correspond to well-test= ed mainline commits with proper Reviewed-by/Tested-by tags. **Positives:** - The dependency chain is correctly ordered (infrastructure first, conversi= on second, followup fixes last). - Each patch has a proper mainline commit reference for traceability. - The series resolves a real syzkaller-found ABBA deadlock between `vblank_= time_lock` and the hrtimer callback. - The code in drm-next confirms these exact changes already landed upstream. **Concerns:** - This is a substantial amount of new infrastructure for a stable backport = (~470 lines added). While the maintainers approved this approach, it carrie= s more regression risk than the minimal fix. - The series was noted as not applying cleanly to drm-next (expected, since= drm-next already has these commits), but the submitter claims it applies c= leanly to 6.18.y =E2=80=94 this should be verified by the stable maintainer= s. - Minor process note: the backporter (Mingyu Wang) does not add their own S= igned-off-by on the cherry-picked patches, which is conventional for `git c= herry-pick -x` but some stable trees expect an explicit sign-off from the s= ubmitter. **Verdict:** The series looks technically correct as a faithful backport of= mainline commits. Acceptable for 6.18.y stable if it applies cleanly there. --- --- Generated by Claude Code Patch Reviewer