From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: Let userspace know about swapped out panthor GEM objects Date: Thu, 23 Apr 2026 08:36:40 +1000 Message-ID: In-Reply-To: <20260421-panthor-bo-reclaim-observability-v2-0-c9135eedfb6f@collabora.com> References: <20260421-panthor-bo-reclaim-observability-v2-0-c9135eedfb6f@collabora.com> X-Mailer: Claude Code Patch Reviewer Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit MIME-Version: 1.0 Overall Series Review Subject: Let userspace know about swapped out panthor GEM objects Author: Nicolas Frattaroli Patches: 5 Reviewed: 2026-04-23T08:36:40.531736 --- This is a clean, well-structured 3-patch series that adds eviction observability for panthor GEM objects through both the standardized fdinfo interface and panthor's debugfs. The design is sound: using a saturating atomic counter to distinguish "evicted to swap" from "never yet backed by pages" is the right approach, and the integration with both the core DRM fdinfo framework (patch 1) and panthor's driver-specific debugfs (patch 2) is straightforward. The main issue is a **missing documentation update**: the new `drm-evicted-` fdinfo key must be documented in `Documentation/gpu/drm-usage-stats.rst`, which defines the standardized userspace-facing interface. Every other memory stat (`total`, `shared`, `resident`, `purgeable`, `active`) is documented there. This is a requirement for any new fdinfo key. Beyond that, the implementation is correct. The ordering of the atomic increment before page cleanup in the eviction path avoids a race window where an object could appear as neither resident nor evicted. The saturation at `INT_MAX` is appropriate. The core DRM code additions are minimal and consistent with the existing pattern. --- --- Generated by Claude Code Patch Reviewer