From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: mm, drm/ttm, drm/xe: Avoid reclaim/eviction loops under fragmentation Date: Tue, 05 May 2026 10:00:00 +1000 Message-ID: In-Reply-To: <20260430191809.2142544-1-matthew.brost@intel.com> References: <20260430191809.2142544-1-matthew.brost@intel.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: mm, drm/ttm, drm/xe: Avoid reclaim/eviction loops under fragmentat= ion Author: Matthew Brost Patches: 14 Reviewed: 2026-05-05T10:00:00.215675 --- This is a 6-patch series (v4) addressing a real and well-documented patholo= gical reclaim loop in Xe GPU systems under memory fragmentation. The proble= m is clearly described: kswapd drives shrinker eviction of GPU BOs, which c= auses rebinds, which trigger more allocation pressure, in a tight loop =E2= =80=94 despite substantial free memory being available (just fragmented). The approach is layered sensibly: two MM core patches (wiring order into sh= rink_control, adding a fragmentation heuristic), two TTM patches (beneficia= l_order reclaim changes, a TTM helper for fragmentation detection), and two= Xe patches (setting the order, using the helper to bail out). **Overall assessment: The series is well-motivated and the approach is reas= onable. However, there are substantive concerns about patch 3's GFP flag ch= anges and some issues with the fragmentation heuristic's placement and logi= c.** --- --- Generated by Claude Code Patch Reviewer