From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: gpu: drm: nova: enable calling into nova-core Date: Thu, 04 Jun 2026 16:23:29 +1000 Message-ID: In-Reply-To: <20260530-nova-exports-v3-0-1202aa339ef7@nvidia.com> References: <20260530-nova-exports-v3-0-1202aa339ef7@nvidia.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: gpu: drm: nova: enable calling into nova-core Author: Alexandre Courbot Patches: 11 Reviewed: 2026-06-04T16:23:29.159435 --- This is a 5-patch series from Alexandre Courbot (NVIDIA) that enables `nova= -drm` to call into `nova-core` by working around the current lack of native= Rust cross-crate module dependency support in the kernel build system. The= approach has three parts: (1) inline some Rust init methods to avoid exces= sively long symbol names, (2) export nova-core's Rust symbols via a generat= ed C file with `EXPORT_SYMBOL_GPL()`, and (3) emit Rust `.rmeta` metadata s= o nova-drm can resolve nova-core's types at compile time. **Strengths:** - Well-documented as a temporary workaround with clear rationale - The `#[inline]` patch (1/5) has standalone merit beyond the workaround - Build system integration is thoughtful (private directory to avoid parall= el build races, order-only dependencies, `filechk` for generated headers) - Patch 5 is honestly labeled as POC/not-for-merge **Concerns:** - Patch 2 exports ALL Rust text symbols from nova-core, which is far more t= han nova-drm needs =E2=80=94 this leaks internal API surface via `EXPORT_SY= MBOL_GPL()` - Patch 3 effectively compiles nova-core twice (once for the .o, once for t= he .rmeta), increasing build time - The patches are based on `drm-rust-next` which has diverged significantly= from the `drm-next` tree (the whole `'bound` lifetime / HRT v5 infrastruct= ure), so they could not be verified against the reference tree **Overall: The series is a reasonable pragmatic workaround. Patches 1-4 loo= k correct for their stated purpose. The POC patch 5 demonstrates sound desi= gn for the inter-module data flow but has some rough edges typical of demo = code.** --- --- Generated by Claude Code Patch Reviewer