From: "Sven Peter" <sven@svenpeter.dev>
To: "Stephen Boyd" <sboyd@kernel.org>,
devicetree@vger.kernel.org, linux-clk <linux-clk@vger.kernel.org>
Cc: linux-arm-kernel@lists.infradead.org,
linux-kernel@vger.kernel.org, "Hector Martin" <marcan@marcan.st>,
"Michael Turquette" <mturquette@baylibre.com>,
"Rob Herring" <robh+dt@kernel.org>,
"Mark Kettenis" <mark.kettenis@xs4all.nl>,
"Arnd Bergmann" <arnd@kernel.org>
Subject: Re: [PATCH 2/3] clk: add support for gate clocks on Apple SoCs
Date: Sun, 30 May 2021 13:17:22 +0200 [thread overview]
Message-ID: <dc4d500e-87f2-47c7-a23d-f0862dc17873@www.fastmail.com> (raw)
In-Reply-To: <162199855335.4130789.17766958356597249549@swboyd.mtv.corp.google.com>
Hi,
Thanks for the review!
On Wed, May 26, 2021, at 05:09, Stephen Boyd wrote:
> Quoting Sven Peter (2021-05-24 11:27:44)
> > Add a simple driver for gate clocks found on Apple SoCs. These don't
> > have any frequency associated with them and are only used to enable
> > access to MMIO regions of various peripherals.
>
> Can we figure out what frequency they are clocking at?
>
I don't think so. I've written some more details about how Apple
uses the clocks in a reply to Rob, but the short version is that
these clock gates are only used as on/off switches. There are
separate clocks that actually have a frequency associated with
them.
> >
> > Signed-off-by: Sven Peter <sven@svenpeter.dev>
> > diff --git a/drivers/clk/Kconfig b/drivers/clk/Kconfig
> > index e80918be8e9c..ac987a8cf318 100644
> > --- a/drivers/clk/Kconfig
> > +++ b/drivers/clk/Kconfig
> > @@ -245,6 +245,18 @@ config CLK_TWL6040
> > McPDM. McPDM module is using the external bit clock on the McPDM bus
> > as functional clock.
> >
> > +config COMMON_CLK_APPLE
> > + tristate "Clock driver for Apple platforms"
> > + depends on ARCH_APPLE && COMMON_CLK
>
> The COMMON_CLK depend is redundant. Please remove.
Removed for v2.
>
> > + default ARCH_APPLE
> > + help
> > + Support for clock gates on Apple SoCs such as the M1.
> > +
> > + These clock gates do not have a frequency associated with them and
> > + are only used to power on/off various peripherals. Generally, a clock
> > + gate needs to be enabled before the respective MMIO region can be
> > + accessed.
> > +
> > config COMMON_CLK_AXI_CLKGEN
> > tristate "AXI clkgen driver"
> > depends on HAS_IOMEM || COMPILE_TEST
> > diff --git a/drivers/clk/clk-apple-gate.c b/drivers/clk/clk-apple-gate.c
> > new file mode 100644
> > index 000000000000..799e9269758f
> > --- /dev/null
> > +++ b/drivers/clk/clk-apple-gate.c
> > @@ -0,0 +1,152 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Apple SoC clock/power gating driver
> > + *
> > + * Copyright The Asahi Linux Contributors
> > + */
> > +
> > +#include <linux/clk.h>
>
> Hopefully this can be droped.
>
> > +#include <linux/clkdev.h>
>
> And this one too. clkdev is not modern.
Okay, will remove both headers (and also fix any code that uses them).
>
> > +#include <linux/err.h>
> > +#include <linux/io.h>
> > +#include <linux/iopoll.h>
> > +#include <linux/clk-provider.h>
> > +#include <linux/of.h>
> > +#include <linux/of_address.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/module.h>
> > +
> > +#define CLOCK_TARGET_MODE_MASK 0x0f
>
> APPLE_ prefix on all these?
Fixed for v2.
>
> > +#define CLOCK_TARGET_MODE(m) (((m)&0xf))
> > +#define CLOCK_ACTUAL_MODE_MASK 0xf0
> > +#define CLOCK_ACTUAL_MODE(m) (((m)&0xf) << 4)
> > +
> > +#define CLOCK_MODE_ENABLE 0xf
> > +#define CLOCK_MODE_DISABLE 0
> > +
> > +#define CLOCK_ENDISABLE_TIMEOUT 100
> > +
> > +struct apple_clk_gate {
> > + struct clk_hw hw;
> > + void __iomem *reg;
> > +};
> > +
> > +#define to_apple_clk_gate(_hw) container_of(_hw, struct apple_clk_gate, hw)
> > +
> > +static int apple_clk_gate_endisable(struct clk_hw *hw, int enable)
> > +{
> > + struct apple_clk_gate *gate = to_apple_clk_gate(hw);
> > + u32 reg;
> > + u32 mode;
> > +
> > + if (enable)
> > + mode = CLOCK_MODE_ENABLE;
> > + else
> > + mode = CLOCK_MODE_DISABLE;
> > +
> > + reg = readl(gate->reg);
> > + reg &= ~CLOCK_TARGET_MODE_MASK;
> > + reg |= CLOCK_TARGET_MODE(mode);
> > + writel(reg, gate->reg);
> > +
> > + return readl_poll_timeout_atomic(gate->reg, reg,
> > + (reg & CLOCK_ACTUAL_MODE_MASK) ==
> > + CLOCK_ACTUAL_MODE(mode),
> > + 1, CLOCK_ENDISABLE_TIMEOUT);
>
> Maybe this
>
> return readl_poll_timeout_atomic(gate->reg, reg,
> (reg & CLOCK_ACTUAL_MODE_MASK) == CLOCK_ACTUAL_MODE(mode),
> 1, CLOCK_ENDISABLE_TIMEOUT);
>
> at the least please don't break the == across two lines.
Ah, sorry. I ran clang-format at the end and didn't catch that line when
I did my final review.
I'll use your suggestion for v2.
>
> > +}
> > +
> > +static int apple_clk_gate_enable(struct clk_hw *hw)
> > +{
> > + return apple_clk_gate_endisable(hw, 1);
> > +}
> > +
> > +static void apple_clk_gate_disable(struct clk_hw *hw)
> > +{
> > + apple_clk_gate_endisable(hw, 0);
> > +}
> > +
> > +static int apple_clk_gate_is_enabled(struct clk_hw *hw)
> > +{
> > + struct apple_clk_gate *gate = to_apple_clk_gate(hw);
> > + u32 reg;
> > +
> > + reg = readl(gate->reg);
> > +
> > + if ((reg & CLOCK_ACTUAL_MODE_MASK) == CLOCK_ACTUAL_MODE(CLOCK_MODE_ENABLE))
>
> Any chance we can use FIELD_GET() and friends? Please don't do bit
> shifting stuff inside conditionals as it just makes life more
> complicated than it needs to be.
Absolutely, thanks for the hint. I didn't know about FIELD_GET and will
use it for v2.
>
> > + return 1;
> > + return 0;
>
> How about just return <logic>?
Good point, fixed for v2.
>
> > +}
> > +
> > +static const struct clk_ops apple_clk_gate_ops = {
> > + .enable = apple_clk_gate_enable,
> > + .disable = apple_clk_gate_disable,
> > + .is_enabled = apple_clk_gate_is_enabled,
> > +};
> > +
> > +static int apple_gate_clk_probe(struct platform_device *pdev)
> > +{
> > + struct device *dev = &pdev->dev;
> > + struct device_node *node = dev->of_node;
> > + const struct clk_parent_data parent_data[] = {
> > + { .index = 0 },
> > + };
>
> Yay thanks for not doing it the old way!
:)
>
> > + struct apple_clk_gate *data;
> > + struct clk_hw *hw;
> > + struct clk_init_data init;
> > + struct resource *res;
> > + int num_parents;
> > + int ret;
> > +
> > + data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
> > + if (!data)
> > + return -ENOMEM;
> > +
> > + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > + data->reg = devm_ioremap_resource(dev, res);
> > + if (IS_ERR(data->reg))
> > + return PTR_ERR(data->reg);
> > +
> > + num_parents = of_clk_get_parent_count(node);
>
> Oh no I spoke too soon.
:(
Sorry, will fix it for v2 to use the new way.
>
> > + if (num_parents > 1) {
> > + dev_err(dev, "clock supports at most one parent\n");
> > + return -EINVAL;
> > + }
> > +
> > + init.name = dev->of_node->name;
> > + init.ops = &apple_clk_gate_ops;
> > + init.flags = 0;
> > + init.parent_names = NULL;
> > + init.parent_data = parent_data;
> > + init.num_parents = num_parents;
> > +
> > + data->hw.init = &init;
> > + hw = &data->hw;
> > +
> > + ret = devm_clk_hw_register(dev, hw);
> > + if (ret)
> > + return ret;
> > +
> > + return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw);
>
> It looks like a one clk per DT node binding which is not how we do it. I
> see Rob has started that discussion on the binding so we can keep that
> there.
>
> > +}
> > +
>
Thanks,
Sven
next prev parent reply other threads:[~2021-05-30 11:17 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-24 18:27 [PATCH 0/3] Apple M1 clock gate driver Sven Peter
2021-05-24 18:27 ` [PATCH 1/3] dt-bindings: clock: add DT bindings for apple,gate-clock Sven Peter
2021-05-24 18:27 ` [PATCH 2/3] clk: add support for gate clocks on Apple SoCs Sven Peter
2021-05-26 3:09 ` Stephen Boyd
2021-05-30 11:17 ` Sven Peter [this message]
2021-05-24 18:27 ` [PATCH 3/3] arm64: apple: add uart gate clocks Sven Peter
2021-05-26 3:10 ` Stephen Boyd
2021-05-30 11:11 ` Sven Peter
2021-05-25 17:41 ` [PATCH 0/3] Apple M1 clock gate driver Rob Herring
2021-05-26 7:18 ` Tony Lindgren
2021-05-30 11:08 ` Sven Peter
2021-06-02 9:26 ` Tony Lindgren
2021-06-03 12:55 ` Sven Peter
2021-06-04 7:43 ` Tony Lindgren
2021-06-05 12:12 ` Sven Peter
2021-06-06 5:59 ` Tony Lindgren
2021-05-30 11:05 ` Sven Peter
2021-06-02 9:28 ` Tony Lindgren
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=dc4d500e-87f2-47c7-a23d-f0862dc17873@www.fastmail.com \
--to=sven@svenpeter.dev \
--cc=arnd@kernel.org \
--cc=devicetree@vger.kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-clk@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=marcan@marcan.st \
--cc=mark.kettenis@xs4all.nl \
--cc=mturquette@baylibre.com \
--cc=robh+dt@kernel.org \
--cc=sboyd@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).