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/amdgpu: fix recursive ww_mutex in devcoredump IB dump Date: Mon, 25 May 2026 19:53:47 +1000 Message-ID: In-Reply-To: <20260520151741.50575-1-mikhail.v.gavrilov@gmail.com> References: <20260520151741.50575-1-mikhail.v.gavrilov@gmail.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: drm/amdgpu: fix recursive ww_mutex in devcoredump IB dump Author: Mikhail Gavrilov Patches: 21 Reviewed: 2026-05-25T19:53:47.734865 --- This mbox contains a patch series evolving across 5 versions (v1 =E2=86=92 = v5), fixing a real lockdep recursive locking splat and potential self-deadl= ock in the amdgpu devcoredump IB dump path. The latest version (v5) is a we= ll-structured 2-patch series that converts `amdgpu_vm_lock_by_pasid()` to u= se `drm_exec` (patch 1) and then uses that to lock all BOs in one ww ticket= (patch 2), as suggested by Christian K=C3=B6nig. **The approach is correct and well-motivated.** The commit messages are tho= rough, the split is logical, and patch 1/2 already has Christian's Reviewed= -by. The series also fixes a pre-existing BO refcount leak on the reserve-f= ailure path. **One correctness bug exists in v5 2/2:** the variable `r` is not initializ= ed before the `drm_exec_until_all_locked` inner for-loop, which means a sta= le value could incorrectly break out of the locking loop when all IB mappin= gs are NULL. This needs a fix before merging. **Review of only the v5 patches follows** (v1=E2=80=93v4 are superseded). --- --- Generated by Claude Code Patch Reviewer