Received-SPF: pass (google.com: domain of linux-kernel+bounces-95659-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240306200910.2732835-1-volodymyr_babchuk@epam.com>
 <CAA8EJppNopEF0DmgjCAJyxe8HRebD26Q8heKKLKbPstdfBOv6A@mail.gmail.com>
 <CAFA6WYMvtNbAMb5bAuH=as01wxACbjs5XcyPrvcSH4YpGGHhpA@mail.gmail.com> <87le6u18xl.fsf@epam.com>
In-Reply-To: <87le6u18xl.fsf@epam.com>
From: Sumit Garg <sumit.garg@linaro.org>
Date: Thu, 7 Mar 2024 19:20:46 +0530
Message-ID: <CAFA6WYOWeBJ+1aT1mRZE_2mNUCBzGGY0qgc_ZTT0dOTAEseGjQ@mail.gmail.com>
Subject: Re: [PATCH] arm64: dts: qcom: sm8150: add reset name for ethernet node
To: Volodymyr Babchuk <Volodymyr_Babchuk@epam.com>
Cc: Dmitry Baryshkov <dmitry.baryshkov@linaro.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>, 
	"linux-arm-msm@vger.kernel.org" <linux-arm-msm@vger.kernel.org>, 
	"devicetree@vger.kernel.org" <devicetree@vger.kernel.org>, 
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"

On Thu, 7 Mar 2024 at 18:43, Volodymyr Babchuk
<Volodymyr_Babchuk@epam.com> wrote:
>
>
> Hello Linux Maintainers, Sumit
>
> First, I am terribly sorry about this half-assed patch. Generally I am
> doing all the required checks. But this change seemed so
> trivial... Anyways, lesson taken, this will not happen anymore.
>
> Sumit Garg <sumit.garg@linaro.org> writes:
>
> > On Thu, 7 Mar 2024 at 12:40, Dmitry Baryshkov
> > <dmitry.baryshkov@linaro.org> wrote:
> >>
> >> On Thu, 7 Mar 2024 at 00:22, Volodymyr Babchuk
> >> <Volodymyr_Babchuk@epam.com> wrote:
> >> >
> >> > Add reset-names property to the ethernet@20000 node. This patch does
> >> > not change behavior on Linux, but it is needed for U-Boot, as it tries
> >> > to find the reset by name, not by index.
> >> >
> >> > Signed-off-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com>
> >> > ---
> >> >  arch/arm64/boot/dts/qcom/sm8150.dtsi | 1 +
> >> >  1 file changed, 1 insertion(+)
> >> >
> >> > diff --git a/arch/arm64/boot/dts/qcom/sm8150.dtsi b/arch/arm64/boot/dts/qcom/sm8150.dtsi
> >> > index 761a6757dc26f..c2e65d6a2ac62 100644
> >> > --- a/arch/arm64/boot/dts/qcom/sm8150.dtsi
> >> > +++ b/arch/arm64/boot/dts/qcom/sm8150.dtsi
> >> > @@ -951,6 +951,7 @@ ethernet: ethernet@20000 {
> >> >
> >> >                         power-domains = <&gcc EMAC_GDSC>;
> >> >                         resets = <&gcc GCC_EMAC_BCR>;
> >> > +                       resets-names = "emac";
> >>
> >> According to the snps,dwmac.yaml schema the "emac" is invalid here.
> >> Only "stmmaceth" and / or "ahb" are permitted here.
> >
> > Okay, it looks like earlier the Linux kernel on Qcom SoCs always
> > assumed that the EMAC reset signal is deserted by prior boot stages.
> > So I suppose we can reuse "stmmaceth" here instead of "emac" with a
> > corresponding change to U-Boot driver as well.
>
> Maybe it would be better to access reset in U-Boot by index, in the
> same way as linux kernel does? I am not sure that "stmmaceth" will be
> correct from the semantic point of view.

I can't see the Linux kernel driver accessing reset by index in this
case (see [1]). It worked for the Qcom case since the reset signal was
by default deserted. IMO, that's an incorrect assumption (you never
know what state the bootloader leaves the reset signal in) to start
with. This should be fixed via explicit MAC reset desertion in the
kernel driver. IOW, this patch would ideally be a fix for the Linux
kernel driver rather than replicating what U-Boot does.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c#n645

>
> As I understand, "stmmac" name is used due to historical reasons in
> Linux, as this driver was introduced for STM SoC initially. But the same
> IP block is being used in many different SoCs made by different vendors
> and there is nothing STM-specific left in it anymore. Especially taking
> into account that this IP-block was designed not by STM but by
> Synopsys/DesignWare.

As DT bindings are the ABI and we have no choice but to live with it.
"stmmaceth" is already being used for "reg-names" and "clock-names"
too.

-Sumit

>
> --
> WBR, Volodymyr