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 and throttling Date: Mon, 25 May 2026 22:14:45 +1000 Message-ID: In-Reply-To: <20260520-feature-dmem-high-v1-1-97ca0cb7f95a@gmail.com> References: <20260520-feature-dmem-high-v1-1-97ca0cb7f95a@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 and throttling Author: Qiliang Yuan Patches: 7 Reviewed: 2026-05-25T22:14:45.117248 --- This single patch adds a `dmem.high` soft limit to the dmem cgroup controll= er, modeled after the memory cgroup's `memory.high`. The idea is sound =E2= =80=94 throttling via return-to-userspace sleep instead of outright allocat= ion failure provides better backpressure for GPU memory. However, the imple= mentation has a **critical sleeping-inside-RCU bug** that will crash the ke= rnel, a **wrong default for the high limit getter**, and several design-lev= el gaps compared to the memcg model it claims to follow (no fast-path guard= , no hierarchy walk for high checking). As written, this patch cannot be ap= plied. --- Generated by Claude Code Patch Reviewer