Linux-Clk Archive on
 help / color / Atom feed
From: Maxime Ripard <>
To: Samuel Holland <>
Cc: Chen-Yu Tsai <>,
	Jassi Brar <>,
	Michael Turquette <>,
	Stephen Boyd <>, Rob Herring <>,
	Mark Rutland <>,
	Corentin Labbe <>,
	Vasily Khoruzhick <>,,,,,
Subject: Re: [PATCH v4 02/10] clk: sunxi-ng: Mark AR100 clocks as critical
Date: Wed, 21 Aug 2019 14:24:36 +0200
Message-ID: <20190821122436.k3s7srhraphfnvgp@flea> (raw)
In-Reply-To: <>

[-- Attachment #1: Type: text/plain, Size: 4574 bytes --]

On Tue, Aug 20, 2019 at 08:02:55AM -0500, Samuel Holland wrote:
> On 8/20/19 2:11 AM, Maxime Ripard wrote:
> > On Mon, Aug 19, 2019 at 10:23:03PM -0500, Samuel Holland wrote:
> >> On sun8i, sun9i, and sun50i SoCs, system suspend/resume support requires
> >> firmware running on the AR100 coprocessor (the "SCP"). Such firmware can
> >> provide additional features, such as thermal monitoring and poweron/off
> >> support for boards without a PMIC.
> >>
> >> Since the AR100 may be running critical firmware, even if Linux does not
> >> know about it or directly interact with it (all requests may go through
> >> an intermediary interface such as PSCI), Linux must not turn off its
> >> clock.
> This paragraph here is the key. The firmware won't necessarily have a device
> tree node, and in the current design it will not, since Linux never communicates
> with it directly. All communication goes through ATF via PSCI.

Sorry, I'm a bit lost in all those ARM firmware interfaces.

I thought SCPI was supposed to be a separate interface that had
nothing to do with PSCI?

Anyway, my main concern is that I don't really want to make exceptions
to the clock handling for everyone's usecase, and this creates a
precedent (and not even a permanent one, since eventually the plan is
to have all the clock handling happening in the firmware, right?).

There's the protected-clocks property in the DT though that will
achieve the same goal. The code to deal with it is not generic at the
moment, but it could be made that way. Would patching the DT to
protect the clock you care about, when you care about protecting them,
be an option for you?

> >> At this time, such power management firmware only exists for the A64 and
> >> H5 SoCs.  However, it makes sense to take care of all CCU drivers now
> >> for consistency, and to ease the transition in the future once firmware
> >> is ported to the other SoCs.
> >>
> >> Leaving the clock running is safe even if no firmware is present, since
> >> the AR100 stays in reset by default. In most cases, the AR100 clock is
> >> kept enabled by Linux anyway, since it is the parent of all APB0 bus
> >> peripherals. This change only prevents Linux from turning off the AR100
> >> clock in the rare case that no peripherals are in use.
> >>
> >> Signed-off-by: Samuel Holland <>
> >
> > So I'm not really sure where you want to go with this.
> >
> > That clock is only useful where you're having a firmware running on
> > the AR100, and that firmware would have a device tree node of its own,
> > where we could list the clocks needed for the firmware to keep
> > running, if it ever runs. If the driver has not been compiled in /
> > loaded, then we don't care either.
> See above. I don't expect that the firmware would have a device tree node,
> because the firmware doesn't need any Linux drivers.
> > But more fundamentally, if we're going to use SCPI, then those clocks
> > will not be handled by that driver anyway, but by the firmware, right?
> In the future, we might use SCPI clocks/sensors/regulators/etc. from Linux, but
> that's not the plan at the moment. Given that it's already been two years since
> I started this project, I'm trying to limit its scope so I can get at least some
> part merged. The first step is to integrate a firmware that provides
> suspend/resume functionality only. That firmware does implement SCPI, and if the
> top-level Linux SCPI driver worked with multiple mailbox channels, it could
> query the firmware's version and fetures. But all of the SCPI commands used for
> suspend/resume must go through ATF via PSCI.

I didn't know that you could / should do that with PSCI / SCPI.

> > So I'm not really sure that we should do it statically this way, and
> > that we should do it at all.
> Do you have a better way to model "firmware uses this clock behind the scenes,
> so Linux please don't touch it"? It's unfortunate that we have Linux and
> firmware fighting over the R_CCU, but since we didn't have firmware (e.g. SCPI
> clocks) in the beginning, it's where we are today.
> The AR100 clock doesn't actually have a gate, and it generally has dependencies
> like R_INTC in use. So as I mentioned in the commit message, the clock will
> normally be on anyway. The goal was to model the fact that there are users of
> this clock that Linux doesn't/can't know about.

Like I said, if that's an option, I'd prefer to have protected-clocks
work for everyone / for sunxi.


Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

  reply index

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-20  3:23 [PATCH v4 00/10] Allwinner sunxi message box support Samuel Holland
2019-08-20  3:23 ` [PATCH v4 01/10] clk: sunxi-ng: Mark msgbox clocks as critical Samuel Holland
2019-08-20  3:23 ` [PATCH v4 02/10] clk: sunxi-ng: Mark AR100 " Samuel Holland
2019-08-20  7:11   ` Maxime Ripard
2019-08-20 13:02     ` Samuel Holland
2019-08-21 12:24       ` Maxime Ripard [this message]
2019-09-05 18:56         ` Stephen Boyd
2019-08-20  3:23 ` [PATCH v4 03/10] dt-bindings: mailbox: Add a sunxi message box binding Samuel Holland
2019-08-20  7:14   ` Maxime Ripard
2019-08-20 13:04     ` Samuel Holland
2019-08-21 12:07       ` Maxime Ripard
2019-08-20  3:23 ` [PATCH v4 04/10] mailbox: sunxi-msgbox: Add a new mailbox driver Samuel Holland
2019-08-20  8:27   ` Maxime Ripard
2019-08-20 11:18   ` Ondřej Jirman
2019-08-20 13:07     ` Samuel Holland
2019-08-20 13:34       ` Ondřej Jirman
2019-08-21 12:30       ` Maxime Ripard
2019-08-20  3:23 ` [PATCH v4 05/10] ARM: dts: sunxi: a80: Add msgbox node Samuel Holland
2019-08-20  8:15   ` Maxime Ripard
2019-08-20 13:17     ` Samuel Holland
2019-08-23 14:56       ` Maxime Ripard
2019-08-20  3:23 ` [PATCH v4 06/10] ARM: dts: sunxi: a83t: " Samuel Holland
2019-08-20  3:23 ` [PATCH v4 07/10] ARM: dts: sunxi: h3/h5: " Samuel Holland
2019-08-20  3:23 ` [PATCH v4 08/10] arm64: dts: allwinner: a64: " Samuel Holland
2019-08-20  3:23 ` [PATCH v4 09/10] arm64: dts: allwinner: h6: " Samuel Holland
2019-08-20  3:23 ` [PATCH v4 10/10] [DO NOT MERGE] drivers: firmware: msgbox demo Samuel Holland
2019-09-09  3:22 ` [PATCH v4 00/10] Allwinner sunxi message box support Ondřej Jirman
2019-09-09  3:54   ` Samuel Holland
2019-09-09 12:36     ` Ondřej Jirman

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:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190821122436.k3s7srhraphfnvgp@flea \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-Clk Archive on

Archives are clonable:
	git clone --mirror linux-clk/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-clk linux-clk/ \
	public-inbox-index linux-clk

Example config snippet for mirrors

Newsgroup available over NNTP:

AGPL code for this site: git clone