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: nova-core: gsp: add continuation record support Date: Thu, 19 Feb 2026 18:43:06 +1000 Message-ID: In-Reply-To: <20260219-cmdq-continuation-v2-0-2e8b7615536f@nvidia.com> References: <20260219-cmdq-continuation-v2-0-2e8b7615536f@nvidia.com> X-Mailer: Claude Code Patch Reviewer Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit MIME-Version: 1.0 Overall Series Review Subject: gpu: nova-core: gsp: add continuation record support Author: Eliot Courtney Patches: 10 Reviewed: 2026-02-19T18:43:06.950624 --- This 9-patch series adds continuation record support to the nova-core GSP command queue. GSP commands larger than 16 pages (64KB) need to be split across multiple queue elements, with the overflow sent as "continuation records." The series takes a well-structured incremental approach: first sorting enum variants, then adding the wait-for-space mechanism, adding size checking, refactoring the send path, and finally introducing the split logic with tests. The overall architecture is sound. The `SplitState`/`SplitCommand` design cleanly separates the splitting logic from the transport layer. Commands that fit in a single element still write directly into the command queue with no extra copies, while oversized commands stage their variable payload into a `KVVec` buffer. The TOCTOU between `driver_write_area_size()` and `driver_write_area()` is safe because the GSP read pointer only advances (growing the writable area) and `&mut self` prevents concurrent driver-side writes. One issue worth discussing is around the `read_poll_timeout` and `driver_write_area` interaction -- see the patch 2 review below. The tests in patch 9 provide good boundary coverage for the split logic. The series is clean and the patches are well-ordered for review. --- Generated by Claude Code Patch Reviewer