From mboxrd@z Thu Jan 1 00:00:00 1970 From: Claude Code Review Bot To: dri-devel-reviews@example.com Subject: Claude review: clk: Add __clk_disable_unprepare_counts_only() and use this in simple[fb|drm] Date: Thu, 28 May 2026 12:34:29 +1000 Message-ID: In-Reply-To: <20260527094811.116977-1-johannes.goede@oss.qualcomm.com> References: <20260527094811.116977-1-johannes.goede@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 Overall Series Review Subject: clk: Add __clk_disable_unprepare_counts_only() and use this in sim= ple[fb|drm] Author: Hans de Goede Patches: 9 Reviewed: 2026-05-28T12:34:29.409298 --- This is a well-motivated series fixing a real problem: simpledrm/simplefb c= all `clk_disable_unprepare()` on teardown, which actually turns off clocks = before the real display driver has loaded, leaving hardware in an undefined= state on Qualcomm platforms. The proposed solution =E2=80=94 decrement ref= counts without touching hardware =E2=80=94 is conceptually sound, and the c= over letter is thorough in discussing alternatives. However, the implementation has a significant gap: the counts-only unprepar= e path still calls `clk_pm_runtime_put()`, which can actually power down th= e clock controller device, undermining the "don't touch hardware" intent. T= he tracepoint emissions in the counts-only path are also misleading. These = issues should be addressed before the clk maintainers are likely to accept = this. The simpledrm error-path change in patch 2 is also questionable =E2=80=94 d= uring init failure, the clocks were genuinely enabled by simpledrm itself a= nd may actually need proper disable/unprepare. --- Generated by Claude Code Patch Reviewer