linux-kernel - Re: [PATCH v7 3/4] clk: samsung: Implement automatic clock gating mode for CMUs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CADrjBPpZY9QfX7YEfM2tkAkt5qGpzYuJ0JkLMK0v4jujO3mRhQ@mail.gmail.com>
Date: Wed, 7 Jan 2026 09:27:37 +0000
From: Peter Griffin <peter.griffin@...aro.org>
To: André Draszik <andre.draszik@...aro.org>
Cc: Krzysztof Kozlowski <krzk+dt@...nel.org>, Alim Akhtar <alim.akhtar@...sung.com>, 
	Tudor Ambarus <tudor.ambarus@...aro.org>, Michael Turquette <mturquette@...libre.com>, 
	Stephen Boyd <sboyd@...nel.org>, Sam Protsenko <semen.protsenko@...aro.org>, 
	Sylwester Nawrocki <s.nawrocki@...sung.com>, Chanwoo Choi <cw00.choi@...sung.com>, 
	Will McVicker <willmcvicker@...gle.com>, Krzysztof Kozlowski <krzk@...nel.org>, devicetree@...r.kernel.org, 
	linux-arm-kernel@...ts.infradead.org, linux-samsung-soc@...r.kernel.org, 
	linux-kernel@...r.kernel.org, linux-clk@...r.kernel.org, 
	kernel-team@...roid.com, Rob Herring <robh@...nel.org>, 
	Conor Dooley <conor+dt@...nel.org>
Subject: Re: [PATCH v7 3/4] clk: samsung: Implement automatic clock gating
 mode for CMUs

Hi André,

On Wed, 7 Jan 2026 at 07:22, André Draszik <andre.draszik@...aro.org> wrote:
>
> Hi,
>
> On Mon, 2025-12-22 at 10:22 +0000, Peter Griffin wrote:
> > Update exynos_arm64_init_clocks() so that it enables the automatic clock
> > mode bits in the CMU option register if the auto_clock_gate flag and
> > option_offset fields are set for the CMU. To ensure compatibility with
> > older DTs (that specified an incorrect CMU reg size), detect this and
> > fallback to manual clock gate mode as the auto clock mode feature depends
> > on registers in this area.
> >
> > The CMU option register bits are global and effect every clock component in
> > the CMU, as such clearing the GATE_ENABLE_HWACG bit and setting GATE_MANUAL
> > bit on every gate register is only required if auto_clock_gate is false.
> >
> > Additionally if auto_clock_gate is enabled the dynamic root clock gating
> > and memclk registers will be configured in the corresponding CMUs sysreg
> > bank. These registers are exposed via syscon, so the register
> > samsung_clk_save/restore paths are updated to also take a regmap.
>
> The implementation described in this paragraph this causes pm-runtime to
> stop working for the CMU if auto clock gating is enabled.
>
> See below.
>
> > [...]
> > diff --git a/drivers/clk/samsung/clk.c b/drivers/clk/samsung/clk.c
> > index c149ca6c221725195faeb76b0d73374c3b48261b..06ea5deef4ee2ffb87dcd14102561886ea80b7bc 100644
> > --- a/drivers/clk/samsung/clk.c
> > +++ b/drivers/clk/samsung/clk.c
> > [...]
> > @@ -357,6 +489,37 @@ void __init samsung_cmu_register_clocks(struct samsung_clk_provider *ctx,
> >               samsung_clk_register_cpu(ctx, cmu->cpu_clks, cmu->nr_cpu_clks);
> >  }
> >
> > +/* Each bit enable/disables DRCG of a bus component */
> > +#define DRCG_EN_MSK  GENMASK(31, 0)
> > +#define MEMCLK_EN    BIT(0)
> > +
> > +/* Enable Dynamic Root Clock Gating (DRCG) of bus components */
> > +void samsung_en_dyn_root_clk_gating(struct device_node *np,
> > +                                 struct samsung_clk_provider *ctx,
> > +                                 const struct samsung_cmu_info *cmu)
> > +{
> > +     if (!ctx->auto_clock_gate)
> > +             return;
> > +
> > +     ctx->sysreg = syscon_regmap_lookup_by_phandle(np, "samsung,sysreg");
>
> With this, the CMU driver (e.g. cmu_hsi0) now gets a reference to the respective
> sysreg (sysreg_hsi0), which in turn creates a regmap with a clock handle
> (CLK_GOUT_HSI0_SYSREG_HSI0_PCLK), and the clock is 'prepared'. Hence the CMU
> providing this clock (cmu_hsi0) is not idle anymore, and runtime PM for this
> CMU can't kick in anymore
>
> I see two straight forward options to fix this:
>
> 1) The easiest fix is to just drop the clock from the sysreg DT node. The sysreg
> clock shouldn't matter now, as the whole CMU is in auto mode now anyway.
>
> 2) change above code to use device_node_to_regmap() which ignores resources (the
> sysreg clock in the example case), and manage the clock manually during save and
> restore in samsung_clk_save() and samsung_clk_restore()

I think option 2 is likely better, as it isn't an ABI break. It would
need a comment explaining why we use that API when a clock resource is
defined in DT for the syscon though (we don't want the clock managed
as it affects runtime PM). Maybe Krzysztof has other thoughts though.

runtime pm isn't currently enabled for gs101, so auto clock support
hasn't regressed anything. But AIUI Andre was close to
posting/enabling it, and this issue would block that.

> Any better ideas?

None I can think of currently.

Thanks,

Peter.