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;
Date:   Tue, 8 Aug 2023 09:27:27 -0500
From:   Andrew Halaney <ahalaney@redhat.com>
To:     Bartosz Golaszewski <brgl@bgdev.pl>
Cc:     Andrew Lunn <andrew@lunn.ch>, Andy Gross <agross@kernel.org>,
        Bjorn Andersson <andersson@kernel.org>,
        Konrad Dybcio <konrad.dybcio@linaro.org>,
        Rob Herring <robh+dt@kernel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
        Conor Dooley <conor+dt@kernel.org>,
        Alex Elder <elder@linaro.org>,
        Srini Kandagatla <srinivas.kandagatla@linaro.org>,
        linux-arm-msm@vger.kernel.org, devicetree@vger.kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Subject: Re: [PATCH 5/9] arm64: dts: qcom: sa8775p-ride: move the reset-gpios
 property of the PHY
Message-ID: <xklkqdodcho4du26mvds4bxrevvwiaftnu7gu2ukchczd7hgcb@ry3dhjnoprgh>
References: <20230807193507.6488-1-brgl@bgdev.pl>
 <20230807193507.6488-6-brgl@bgdev.pl>
 <siqiyihftz3musfjulpcqunhgi7npftumrfwfyh2pqnlx6zeb7@rrpwmkvjshfb>
 <da679b5e-6712-4849-b29c-6aa42022abc4@lunn.ch>
 <ld2j4llgfba6j43gesqxs6wz2baucka5scbj4nef5ehbex2cmt@d4dxsqp2vuoj>
 <CAMRc=MdLky5sUbdFGFc+as906kr-J_XDmKmYtBBCHvETvqtAQA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <CAMRc=MdLky5sUbdFGFc+as906kr-J_XDmKmYtBBCHvETvqtAQA@mail.gmail.com>
Precedence: bulk

On Tue, Aug 08, 2023 at 02:16:50PM +0200, Bartosz Golaszewski wrote:
> On Tue, Aug 8, 2023 at 12:27 AM Andrew Halaney <ahalaney@redhat.com> wrote:
> >
> > On Mon, Aug 07, 2023 at 11:51:40PM +0200, Andrew Lunn wrote:
> > > > > I have proposed a solution for this problem in 2020 but it never got
> > > > > upstream. Now we have a workaround in place which allows us to hard-code
> > > > > the PHY id in the compatible property, thus skipping the ID scanning).
> > > >
> > > > nitpicky, but I think that already existed at that time :D
> > >
> > > Yes, it has been there are long long time. It is however only in the
> > > last 5 years of so has it been seen as a solution to the chicken egg
> > > problem.
> > >
> > > > >           sgmii_phy: phy@8 {
> > > > > +                 compatible = "ethernet-phy-id0141.0dd4";
> > > > >                   reg = <0x8>;
> > > > >                   device_type = "ethernet-phy";
> > > > > +                 reset-gpios = <&pmm8654au_2_gpios 8 GPIO_ACTIVE_LOW>;
> > > > > +                 reset-deassert-us = <70000>;
> > > >
> > > > Doesn't this need reset-assert-us?
> > >
> > > If i remember correctly, there is a default value if DT does not
> > > provide one.
> > >
> >
> > I've been trying to make sure I view devicetree properties as an OS
> > agnostic ABI lately, with that in mind...
> >
> > The dt-binding says this for ethernet-phy:
> >
> >   reset-assert-us:
> >     description:
> >       Delay after the reset was asserted in microseconds. If this
> >       property is missing the delay will be skipped.
> >
> > If the hardware needs a delay I think we should encode it based on that
> > description, else we risk it starting to look like a unit impulse!
> >
> 
> Please note that the mdio-level delay properties are not the same as
> the ones on the PHY levels.
> 
> reset-delay-us - this is the delay BEFORE *DEASSERTING* the reset line
> reset-post-delay-us - this is the delay AFTER *DEASSERTING* the reset line
> 
> On PHY level we have:
> 
> reset-assert-us - AFTER *ASSERTING*
> reset-deassert-us - AFTER *DEASSERTING*
> 
> There never has been any reset-assert delay on that line before. It
> doesn't look like we need a delay BEFORE deasserting the line, do we?
> 

The MDIO reset-delay-us happens "before deassert"/"after assert", so to
make things a proper move here I think it needs a resrt-assert, otherwise
behavior with respect to reset timing is definitely changed from this
patch!

Here's a trimmed version of the reset handling in mdio_bus.c to back
that up:

	/* assert bus level PHY GPIO reset */
	gpiod = devm_gpiod_get_optional(&bus->dev, "reset", GPIOD_OUT_HIGH);
    ...
	} else	if (gpiod) {
		bus->reset_gpiod = gpiod;
		fsleep(bus->reset_delay_us);
		gpiod_set_value_cansleep(gpiod, 0);
		if (bus->reset_post_delay_us > 0)
			fsleep(bus->reset_post_delay_us);
	}

so its assert reset, sleep reset_delay_us, deassert, sleep
reset_post_delay_us for the MDIO case.

Thanks,
Andrew