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/v3d: Fix perfmon locking and cross-queue isolation Date: Thu, 04 Jun 2026 14:44:42 +1000 Message-ID: In-Reply-To: <20260531-v3d-perfmon-lifetime-v2-0-60ed4485a203@igalia.com> References: <20260531-v3d-perfmon-lifetime-v2-0-60ed4485a203@igalia.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/v3d: Fix perfmon locking and cross-queue isolation Author: =3D?utf-8?q?Ma=3DC3=3DADra_Canal?=3D Patches: 10 Reviewed: 2026-06-04T14:44:42.323803 --- This is a well-structured 4-patch series from Ma=C3=ADra Canal that address= es three real, long-standing correctness issues in v3d's performance monito= r subsystem: (1) lock-free access to `active_perfmon`, (2) perfmons never b= eing stopped after job completion, and (3) cross-queue counter pollution. T= he cover letter clearly describes the problems and the approach. The series is logically decomposed: patch 1 is a self-contained stable fix = for refcount leaks, patch 2 is the major locking refactor, patch 3 adds cro= ss-queue serialization via fence dependencies, and patch 4 is a trivial cle= anup. The dependency on a prerequisite series (the submission refactoring) = introduces a `v3d_submit` abstraction used in patches 3-4 that isn't presen= t in drm-next today, but the patches are internally consistent. **Overall verdict**: Good series. A few points worth discussing, mainly aro= und locking subtleties in patch 2 and a potential race window in patch 3, b= ut nothing blocking. --- --- Generated by Claude Code Patch Reviewer