From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: Introduce cold reset recovery method Date: Sun, 12 Apr 2026 14:32:33 +1000 Message-ID: In-Reply-To: <20260406142325.157035-6-mallesh.koujalagi@intel.com> References: <20260406142325.157035-6-mallesh.koujalagi@intel.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: Introduce cold reset recovery method Author: Mallesh Koujalagi Patches: 8 Reviewed: 2026-04-12T14:32:33.157016 --- This v3 series adds cold-reset recovery support for the Xe driver, building= on top of an unmerged RAS/PCI-error-handling infrastructure series ([PATCH= ] "Introduce Xe Uncorrectable Error Handling"). The series introduces `DRM_= WEDGE_RECOVERY_COLD_RESET` at the DRM core level and wires it up for PUNIT = errors in the Xe driver. **Key concerns:** 1. **Patch 1 is marked "DO NOT REVIEW" but carries significant architectura= l decisions** that directly affect the reviewability of patches 2-4. The PC= I error recovery path in `xe_pci_error_slot_reset()` has serious issues (ca= lling `pdev->driver->remove()`/`probe()` directly, using `devres_release_gr= oup()` in a fragile way) that affect the entire series. The dependency chai= n also means the `xe_sysctrl_mailbox` infrastructure (`has_sysctrl`, `xe_sy= sctrl_send_command`, etc.) doesn't exist in drm-next, making it impossible = to verify compilation or correctness. 2. **The `in_recovery` atomic overlaps with the `wedged.flag` atomic** in u= nclear ways. Both serve to block operations during error states, but the in= teraction and lifecycle differences are not well documented. 3. **The cold-reset recovery script in the documentation is incomplete** = =E2=80=94 it only handles the slot-power-cycle case and does nothing meanin= gful when no slot is found (just prints "No slot found" and exits). --- --- Generated by Claude Code Patch Reviewer