Date:   Wed, 4 Aug 2021 19:09:46 -0700
From:   Tao Ren <rentao.bupt@...il.com>
To:     Joel Stanley <joel@....id.au>
Cc:     Rob Herring <robh+dt@...nel.org>, Andrew Jeffery <andrew@...id.au>,
        devicetree <devicetree@...r.kernel.org>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>,
        linux-aspeed <linux-aspeed@...ts.ozlabs.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        OpenBMC Maillist <openbmc@...ts.ozlabs.org>,
        Tao Ren <taoren@...com>
Subject: Re: [PATCH 6/6] ARM: dts: aspeed: Add Facebook Fuji (AST2600) BMC

On Thu, Aug 05, 2021 at 01:37:29AM +0000, Joel Stanley wrote:
> On Thu, 5 Aug 2021 at 01:20, Tao Ren <rentao.bupt@...il.com> wrote:
> >
> > On Thu, Aug 05, 2021 at 12:28:02AM +0000, Joel Stanley wrote:
> > > On Wed, 28 Jul 2021 at 23:38, <rentao.bupt@...il.com> wrote:
> > > >
> > > > From: Tao Ren <rentao.bupt@...il.com>
> > > >
> > > > Add initial version of device tree for Facebook Fuji (AST2600) BMC.
> > >
> > > I like to read what kind of platform the BMC is going into if you can
> > > add that detail, but it's not essential.
> >
> > Sure. I will add more details in v2.
> >
> > > > +&spi1 {
> > > > +       status = "okay";
> > > > +
> > > > +       /*
> > > > +        * Customize spi1 flash memory size to 32MB (maximum flash size on
> > > > +        * the bus) to save vmalloc space.
> > > > +        */
> > > > +       reg = < 0x1e630000 0xc4
> > > > +               0x30000000 0x2000000 >;
> > >
> > > Which driver supports this?
> > >
> > > It would be great to see Facebook work to get the SPI NOR driver for
> > > the ast2600 merged to mainline.
> > >
> > > I doubt the IBM team will get to this, as we are using eMMC instead.
> >
> > Ah, I just checked aspeed-g6.dtsi (ast2600-spi) in mainline and I thought
> > the driver patches were also upstreamed. Let me remove the entry for now,
> > and will add it back when the driver is ready.
> 
> If it helps you, I don't mind it staying in your dts. I would like to
> see some effort on the driver though.
> 
> >
> > >
> > > > +
> > > > +       flash@0 {
> > > > +               status = "okay";
> > > > +               m25p,fast-read;
> > > > +               label = "spi1.0";
> > > > +               spi-max-frequency = <5000000>;
> > > > +
> > > > +               partitions {
> > > > +                       compatible = "fixed-partitions";
> > > > +                       #address-cells = <1>;
> > > > +                       #size-cells = <1>;
> > > > +
> > > > +                       flash1@0 {
> > > > +                               reg = <0x0 0x2000000>;
> > > > +                               label = "system-flash";
> > > > +                       };
> > > > +               };
> > > > +       };
> > > > +};
> > >
> > > > +&ehci1 {
> > > > +       status = "okay";
> > > > +};
> > >
> > > Have you verified that USB works with mainline? I've had reports of it
> > > working on 5.8 but it seems to have regressed as of v5.10.
> >
> > It stopped working on ASPEED since commit 280a9045bb18 ("ehci: fix EHCI
> > host controller initialization sequence"): ehci_handshake() returns
> > error because HCHalted bit EHCI24[12] stays at 1.
> >
> > I have a dirty hack in my tree (ignoring the halt bit) and it "works" on
> > AST2500 and AST2600. Let me send an email to openbmc and aspeed email
> > groups to see if anyone has more context.
> 
> We discovered the same, except in my testing it wasn't actually
> "working" on my EVB. I have a A3 EVB on my desk and when plugging in a
> USB key nothing happened.

We have a non-removable USB device on Fuji (AST2600) BMC and it can be
enumerated successfully (after applying the hack). Let me vist the lab
and plug a USB key to CMM (AST2500) BMC sometime next week. Will collect
results and send out email then.

> 
> >
> > Meanwhile, should I delete the entry until the fix is ready in driver?
> 
> Again, leave it in I think.
> 
> >
> > > > +&mdio1 {
> > > > +       status = "okay";
> > > > +
> > > > +       ethphy3: ethernet-phy@13 {
> > > > +               compatible = "ethernet-phy-ieee802.3-c22";
> > > > +               reg = <0x0d>;
> > > > +       };
> > > > +};
> > > > +
> > > > +&mac3 {
> > > > +       phy-handle = <&ethphy3>;
> > >
> > > status = okay?
> > >
> > > You should specify the pinmux too I think, even if the default happens
> > > to work, so that other devices cannot claim the pins.
> >
> > status is set in ast2600 common dtsi, but let me set it here to avoid
> > confusion. Will update pinmux in v2. Thanks.
> 
> Are you looking at mainline? I think it's disabled by default there:
> 
> https://elixir.bootlin.com/linux/latest/source/arch/arm/boot/dts/aspeed-g6.dtsi#L246

Sorry for the confusion, Joel. I mean the device is enabled in
"ast2600-facebook-netbmc-common.dtsi" (patch #3 in the series). Do we
still need to set status to "okay" explicitly in platform dts (to avoid
confusion)?

&mac3 {
        status = "okay";
        phy-mode = "rgmii";
        pinctrl-names = "default";
        pinctrl-0 = <&pinctrl_rgmii4_default>;
};


> 
> >
> > >
> > > Cheers,
> > >
> > > Joel
> >
> > Cheers,
> >
> > Tao

Powered by blists - more mailing lists