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/msm/a6xx: Fix gpu init from secure world Date: Wed, 25 Mar 2026 07:32:10 +1000 Message-ID: In-Reply-To: <20260324-a8xx-gpu-batch2-v1-10-fc95b8d9c017@oss.qualcomm.com> References: <20260324-a8xx-gpu-batch2-v1-0-fc95b8d9c017@oss.qualcomm.com> <20260324-a8xx-gpu-batch2-v1-10-fc95b8d9c017@oss.qualcomm.com> X-Mailer: Claude Code Patch Reviewer Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Patch Review Moves the SCM call from probe time to `a6xx_gmu_resume()`. The rationale is= that GDSC/clocks must be on for the SCM call. **Concern:** This means the SCM call now happens on every resume, not just = once at probe. The `dev_warn_once` in the SCM-unavailable path suggests the= original intent was one-time init. If the SCM call is idempotent this is f= ine, but it adds resume latency. Also, `has_ray_tracing` gets set on every = resume =E2=80=94 if something could change it between suspend/resume, that'= s intentional; otherwise it's wasted work. **Nit:** The error message has a double space: `"SCM call failed\n"`. **Missing error handling:** If `a6xx_gmu_secure_init()` fails during resume= , the code returns early without cleaning up the GMU state that was already= initialized in the preceding steps of `a6xx_gmu_resume()`. --- Generated by Claude Code Patch Reviewer