[U-Boot] dm: pinctrl: Prevent (re-)configuring pins when already done before relocation

[Simon Glass]

Hi Lukasz,

On Sat, 12 Jan 2019 at 14:46, Lukasz Majewski <lukma at denx.de> wrote:
>
> Hi Alex,
>
> > On Tue, Jan 8, 2019 at 11:24 PM Lukasz Majewski <lukma at denx.de> wrote:
> > >
> > > Hi Alex,
> > >
> > > > On Tue, Jan 8, 2019 at 10:52 PM Lukasz Majewski <lukma at denx.de>
> > > > wrote:
> > > > >
> > > > > Hi Alex,
> > > > >
> > > > > > On Thu, Dec 27, 2018 at 3:49 PM Tom Rini <trini at konsulko.com>
> > > > > > wrote:
> > > > > > >
> > > > > > > On Tue, Dec 18, 2018 at 12:30:50PM +0100, Lukasz Majewski
> > > > > > > wrote:
> > > > > > > > This commit prevents from re-configuring pins if those
> > > > > > > > were configured before relocation.
> > > > > > > >
> > > > > > > > Some pins - like UART or DDR must be setup before
> > > > > > > > relocation (as they have 'u-boot,dm-pre-reloc' property
> > > > > > > > set in DTS). Without this change, those pins are
> > > > > > > > re-configured after relocation (pre_reloc_only = 0, so we
> > > > > > > > do not "continue"). Such behavior may be a problem for
> > > > > > > > DDR PAD configuration, as they might be already
> > > > > > > > leveled/tuned with original setup).
> > > > > > > >
> > > > > > > > Signed-off-by: Lukasz Majewski <lukma at denx.de>
> > > > > > >
> > > > > > > Applied to u-boot/master, thanks!
> > > > > > >
> > > > > >
> > > > > > I've bisected out to this commit and it's slightly broken
> > > > > > things for me on an AM3352. It all works fine so long as I
> > > > > > boot MLO from MMC (so the MMC is probed, pinctrl setup), but
> > > > > > if I boot from UART then I get to full U-Boot, MMC hasn't
> > > > > > been probed and the pinmuxing isn't set up for the MMC.
> > > > >
> > > > > I suppose that the pinmux node have set "u-boot,dm-pre-reloc"
> > > > > property?
> > > > >
> > > >
> > > > Yeah, well "u-boot,dm-spl", though now I re-read the docs I guess
> > > > that's equivalent to "u-boot,dm-pre-reloc" as I've no TPL.
> > >
> > > I do need to check this -> as in my case I do have
> > > "u-boot,dm-pre-reloc" property. And yes, in my case I do not have
> > > TPL too (just SPL).
> > >
> > > >
> > > > > The problem is not with lack of eMMC probing - it is with
> > > > > pinctrl nodes having "u-boot,dm-pre-reloc" set in DTS and the
> > > > > eMMC is probed after MLO/SPL.
> > > > > It looks like your use case implicitly depends on pinmux being
> > > > > reconfigured no matter if we are pre-relocated (MLO) or
> > > > > afterwards.
> > > >
> > > > Yes, that's a much clearer description...
> > >
> > > Frankly, your mail opened my eyes to a use case which was handled (I
> > > assume on purpose) with the old code - the "cross" boot with pinmux
> > > and devices.
> > >
> > > For example boot from UART (MLO/SPL) and then continue with eMMC
> > > (which require pinmux).
> > >
> >
> > So I guess this needs something like a bloblist to communicate the
> > initialisation state between MLO/SPL and u-boot,
>
> Yes. IMHO this is one way to do it. There is already a
> DM_FLAG_ACTIVATED, which indicates if the device was probed. However,
> it is not "dragged" from MLO / SPL to u-boot.img
>
> However, I think that we shall introduce a new flag -
> DM_FLAG_CONFIGURE_ONCE [*], which would indicate that the device need
> single configuration.
>
> This flag could be used with DM_PINMUX driver, to avoid
> re-configuration.
>
> The idea would be to use bloblist (only when [*] is set) to pass tuples:
> (driver_name, config_status)

That seems reasonable, but I don't like the idea of using a name.
Could we have some sort of numeric value instead?

>
> and in the core of DM just check if we shall re-configure the driver or
> not.
>
> > which I've not the
> > faintest idea where to start on.
> >
> > Horrible as they are, is a CONFIG symbol a possibility whilst we
> > figure out how to solve this properly?
>
> The bloblist can be activated with CONFIG_BLOBLIST
>
> >
> > I guess another option would be an additional two properties that give
> > the two behaviours, but that feels like something we don't want long
> > term.
>
> As we strive to use as little "u-boot,XXXXXX" extra properties to avoid
> problems with Linux re-sync, I think that we shall use bloblist.

Regards,
Simon