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: tyr: use lifetime-bound IoMem Date: Wed, 27 May 2026 15:26:02 +1000 Message-ID: In-Reply-To: <20260525230152.277820-1-dakr@kernel.org> References: <20260525230152.277820-1-dakr@kernel.org> 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: tyr: use lifetime-bound IoMem Author: Danilo Krummrich Patches: 7 Reviewed: 2026-05-27T15:26:02.455706 --- This is a clean 2-patch refactoring series from Danilo Krummrich that moder= nizes the tyr (Arm Mali Valhall) DRM driver to use the new driver-core life= time infrastructure. The series depends on a prerequisite patch set [1] tha= t adds lifetime-bound driver data to the platform driver framework, and is = based on drm-rust-next (not the drm-next tree available here =E2=80=94 henc= e the apply failure and the substantial differences in the register access = pattern). **Patch 1** separates the `platform::Driver` trait implementation onto a ne= w zero-sized `TyrPlatformDriver` type, keeping `TyrPlatformDriverData` as t= he associated `Data<'bound>` type. This is a standard pattern enabling the = new GAT-based lifetime-bound driver data. **Patch 2** takes advantage of that lifetime binding to use `IoMem<'a>` dir= ectly on the stack in probe, removing the `Devres`/`Arc` wrapping that was = previously needed to satisfy `'static` requirements. This simplifies regist= er access by eliminating the `(*iomem).access(dev)?` indirection. The overall direction is good =E2=80=94 the I/O memory is only used during = probe and never stored in driver data, so the `Devres` revocation machinery= was unnecessary overhead. One minor issue noted below. --- Generated by Claude Code Patch Reviewer