[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ZZU0eiw_1qSOklRN@hovoldconsulting.com>
Date: Wed, 3 Jan 2024 11:18:34 +0100
From: Johan Hovold <johan@...nel.org>
To: Konrad Dybcio <konrad.dybcio@...aro.org>
Cc: Bjorn Andersson <andersson@...nel.org>, Andy Gross <agross@...nel.org>,
Michael Turquette <mturquette@...libre.com>,
Stephen Boyd <sboyd@...nel.org>, Rob Herring <robh+dt@...nel.org>,
Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
Conor Dooley <conor+dt@...nel.org>,
Marijn Suijten <marijn.suijten@...ainline.org>,
linux-arm-msm@...r.kernel.org, linux-clk@...r.kernel.org,
linux-kernel@...r.kernel.org, devicetree@...r.kernel.org,
Bryan O'Donoghue <bryan.odonoghue@...aro.org>
Subject: Re: [PATCH v4 02/12] clk: qcom: Use qcom_branch_set_clk_en()
On Tue, Jan 02, 2024 at 03:27:29PM +0100, Konrad Dybcio wrote:
> On 2.01.2024 11:35, Johan Hovold wrote:
> > On Sat, Dec 30, 2023 at 02:04:04PM +0100, Konrad Dybcio wrote:
> >> @@ -3010,10 +3010,8 @@ static int camcc_sc8280xp_probe(struct platform_device *pdev)
> >> clk_lucid_pll_configure(&camcc_pll6, regmap, &camcc_pll6_config);
> >> clk_lucid_pll_configure(&camcc_pll7, regmap, &camcc_pll7_config);
> >>
> >> - /*
> >> - * Keep camcc_gdsc_clk always enabled:
> >> - */
> >> - regmap_update_bits(regmap, 0xc1e4, BIT(0), 1);
> >> + /* Keep the critical clocks always-on */
> >> + qcom_branch_set_clk_en(regmap, 0xc1e4); /* CAMCC_GDSC_CLK */
> >
> > I still think something along the lines of
> >
> > /* Keep some clocks always on */
> >
> > is preferred as it is far from obvious why a camera clock would be
> > considered "critical".
> >
> > Or perhaps you can come up with a better description of why we've
> > decided not to model these clocks and just leave them ungated.
> Technically they're not really super critical if the hardware is
> not in use.. It's just that at one point Qualcomm decided to take
> the lazy decision to keep them always-on downstream and we seem to
> have agreed on going with that, instead of pm_clk (remember my old
> attempt at getting rid of this on dispcc-sc8280xp?)..
>
> For now, I was just trying to clean this up a bit before looking
> into a better solution for this (probably a whole lot of pm_clks
> with some clever handle-getting due to different ways of grabbing
> clock sources.. by-name vs by-index vs global lookup that we've
> accumulated over the years).
Yeah, that's fine. I'm not saying you have to come up with a better way
of describing these for this series, but I find calling them "critical"
throughout is more confusing than the current unspecified comments about
leaving some clocks on (i.e. without a proper motivation or even hint
about why they are being kept always on).
Johan
Powered by blists - more mailing lists