Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <0dde6d95f4e92da68ad3a3122056bfbd.sboyd@kernel.org>
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
In-Reply-To: <1jpm3xoea9.fsf@starbuckisacylon.baylibre.com>
References: <20230731042807.1322972-1-wenst@chromium.org> <1jpm3xoea9.fsf@starbuckisacylon.baylibre.com>
Subject: Re: [PATCH] clk: meson: meson8b: Simplify notifier clock lookup
From:   Stephen Boyd <sboyd@kernel.org>
Cc:     linux-amlogic@lists.infradead.org, linux-clk@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org
To:     Chen-Yu Tsai <wenst@chromium.org>,
        Jerome Brunet <jbrunet@baylibre.com>,
        Kevin Hilman <khilman@baylibre.com>,
        Martin Blumenstingl <martin.blumenstingl@googlemail.com>,
        Neil Armstrong <neil.armstrong@linaro.org>
Date:   Wed, 09 Aug 2023 14:55:40 -0700
User-Agent: alot/0.10
Precedence: bulk

Quoting Jerome Brunet (2023-08-08 06:21:46)
>=20
> On Mon 31 Jul 2023 at 12:27, Chen-Yu Tsai <wenst@chromium.org> wrote:
> > @@ -3847,9 +3846,7 @@ static void __init meson8b_clkc_init_common(struc=
t device_node *np,
> >        * tricky programming sequence will be handled by the forthcoming
> >        * coordinated clock rates mechanism once that feature is release=
d.
> >        */
> > -     notifier_clk_name =3D clk_hw_get_name(&meson8b_cpu_scale_out_sel.=
hw);
> > -     notifier_clk =3D __clk_lookup(notifier_clk_name);
> > -     ret =3D clk_notifier_register(notifier_clk, &meson8b_cpu_nb_data.=
nb);
> > +     ret =3D clk_notifier_register(meson8b_cpu_scale_out_sel.hw.clk, &=
meson8b_cpu_nb_data.nb);
>=20
> Hi Chen-Yu,
>=20
> Your patch seems valid, as CCF stands right now.
>=20
> However, I believe there is a will to drop the 'struct clk' instance that
> automatically gets created with each 'struct clk_hw'. This change would
> not help going in this direction
>=20

Yes. Use clk_hw_get_clk() here, or introduce a
clk_hw_notifier_register() API that does that automatically. We will
have to put the clk on unregister path as well though, so maybe struct
clk_notifier needs to be extended to store the clk_hw pointer. The
notifier code kinda needs an overhaul to be clk_hw/clk_core aware. I
think it predates the introduction of struct clk_core? It's a bunch of
work I've been putting off.