From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: cgroup/dmem: implement dmem.high soft limit via prioritized eviction Date: Thu, 04 Jun 2026 15:52:23 +1000 Message-ID: In-Reply-To: <20260530-feature-dmem-high-v4-1-ee7c6ec1c8da@gmail.com> References: <20260530-feature-dmem-high-v4-1-ee7c6ec1c8da@gmail.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: cgroup/dmem: implement dmem.high soft limit via prioritized evicti= on Author: Qiliang Yuan Patches: 1 Reviewed: 2026-06-04T15:52:23.141901 --- This is a single-patch series (v4) adding a `dmem.high` soft limit to the d= mem cgroup v2 controller. The concept is sound: when eviction is triggered = (e.g., by a `max` limit hit), BOs from cgroups exceeding their `dmem.high` = soft limit are evicted first, providing a prioritized reclaim mechanism rat= her than the rejected sleep-on-allocation approach from v1. The cgroupfs plumbing (control file, get/set, initialization) follows exist= ing patterns well. However, the TTM eviction changes have two significant b= ugs: Pass 2 runs with blocking lock instead of trylock (the `trylock_only` = flag is never restored after Pass 1), and the `try_high` path in `dmem_cgro= up_state_evict_valuable()` skips the hierarchy check between `limit_pool` a= nd `test_pool`, meaning BOs from completely unrelated cgroups can be evicte= d if any of their ancestors happens to exceed its high limit. **Verdict: Needs revision** =E2=80=94 the trylock regression and the missin= g hierarchy check need to be fixed before this can be merged. --- --- Generated by Claude Code Patch Reviewer