From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: Revert "drm/imagination: Warn or error on unsupported hardware" Date: Sat, 16 May 2026 15:12:53 +1000 Message-ID: In-Reply-To: References: 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: Revert "drm/imagination: Warn or error on unsupported hardware" Author: Geert Uytterhoeven Patches: 4 Reviewed: 2026-05-16T15:12:53.518717 --- This is a single patch that reverts commit `1c21f240fbc1e47b` ("drm/imagina= tion: Warn or error on unsupported hardware"). That commit introduced a GPU= support allowlist that blocks driver probing on GPU cores not explicitly l= isted as `PVR_GPU_SUPPORTED` or `PVR_GPU_EXPERIMENTAL`, unless the user set= s the `exp_hw_support` module parameter. The revert is motivated by the fact that Renesas R-Car SoCs (which have Pow= erVR Rogue GPUs like GX6250 and GE7800) are blocked by this allowlist =E2= =80=94 their BVNC values are not in the switch statement, so they fall into= `PVR_GPU_UNKNOWN` and the driver refuses to probe with `-ENODEV`. **Assessment:** The patch is technically correct as a revert =E2=80=94 it c= leanly removes the support-level checking infrastructure and replaces the c= all with `return 0`. The commit message is well-documented with extensive r= eferences to DT bindings, DTS additions, firmware availability, and the pow= er domain fix that unblocked board enablement. However, the better fix is arguably **not** a full revert but rather adding= the Renesas GPU BVNC values to the allowlist. The support-level mechanism = was intentionally introduced as a safety gate for the early-stage driver. R= everting it entirely removes protection for truly unknown/untested hardware= configurations. That said, the commit message makes a valid argument that = the allowlist approach creates ongoing friction for new platform enablement= =E2=80=94 every new SoC with a PowerVR GPU must also have its BVNC added t= o the driver's allowlist, which is maintenance burden that doesn't exist in= other DRM drivers. **Recommendation:** This is a policy/design decision for the drm/imaginatio= n maintainers. The patch is correct in what it does, but the maintainers sh= ould weigh whether to accept the full revert vs. expanding the allowlist. T= he author has clearly thought about this and chosen the revert deliberately. --- Generated by Claude Code Patch Reviewer