From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: Update 'none' recovery method for DRM wedged event Date: Wed, 04 Mar 2026 07:31:27 +1000 Message-ID: In-Reply-To: <20260303110556.3577612-1-raag.jadav@intel.com> References: <20260303110556.3577612-1-raag.jadav@intel.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: Update 'none' recovery method for DRM wedged event Author: Raag Jadav Patches: 7 Reviewed: 2026-03-04T07:31:27.271209 --- This 3-patch series broadens the meaning of `WEDGED=none` to cover a new use case: the Xe driver's `XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET` debug mode, where the device is intentionally left wedged (no GT reset) so developers can inspect the hang state. The core concern is a **UAPI semantic change**. The current documentation and log message for `WEDGED=none` imply the device was already recovered (via reset) and is functional again. This series repurposes it to also mean "device is permanently wedged, don't try to recover it" for the debug case. These are fundamentally different device states from userspace's perspective: in one case the device works; in the other, all IOCTLs are blocked and only a rebind can restore it. Any existing userspace consumer processing `WEDGED=none` as "device is healthy, carry on" would be broken by the new usage. The series would benefit from clearer documentation that distinguishes these two scenarios, or alternatively, considering whether a different mechanism (perhaps `WEDGED=vendor-specific` or a new method) would better express the "permanently wedged for debug" semantic without changing the existing contract. --- Generated by Claude Code Patch Reviewer