[PATCH] sunxi: Bring back SD card as MMC device 0

[Andre Przywara]

On Mon, 26 Apr 2021 17:52:51 +0530
Jagan Teki <jagan at amarulasolutions.com> wrote:

Hi Jagan,

> On Mon, Apr 26, 2021 at 4:57 PM Andre Przywara <andre.przywara at arm.com> wrote:
> >
> > On Sun, 25 Apr 2021 18:03:05 +0530
> > Jagan Teki <jagan at amarulasolutions.com> wrote:
> >
> > Hi Jagan,
> >
> > thanks for your input!
> >  
> > > On Sun, Apr 25, 2021 at 3:30 PM Andre Przywara <andre.przywara at arm.com> wrote:  
> > > >
> > > > On Fri, 16 Apr 2021 12:08:09 +0100
> > > > Andre Przywara <andre.przywara at arm.com> wrote:
> > > >
> > > > Hi,
> > > >  
> > > > > Commit 2243d19e5618 ("mmc: mmc-uclass: Use dev_seq() to read aliases
> > > > > node's index") now actually enforces U-Boot's device enumeration policy,
> > > > > where explicitly named devices come first, then any other non-named
> > > > > devices follow, without filling gaps.
> > > > >
> > > > > For quite a while we have had an "mmc1 = &mmc2;" alias in our
> > > > > sunxi-u-boot.dtsi, which now leads to the problem that the SD card
> > > > > (which was always mmc device 0) now gets to be number 2.
> > > > > I guess this breaks quite some boot scripts, and is confusing at least.
> > > > >
> > > > > Just add an explicit mmc0 alias in the very same file to fix this and
> > > > > restore the old behaviour.  
> > > >
> > > > Can someone please say if this is the right solution?
> > > > I think the SD card has always been mmc device 0 in U-Boot, so I think
> > > > it's worth keeping that. Just not sure if this is the right way of
> > > > fixing that?  
> > >
> > > Playing with aliases always gets confused and might get a different
> > > behavior, IMHO.  
> >
> > I am not so sure about that, since aliases force a fixed numbering, so
> > it should be less confusing. We have the same problem in the kernel
> > now, and Samuel sent some patches to have aliases in the mainline DTs
> > as well [1].  
> 
> Okay.
> 
> >  
> > > Detect the dev number by U-Boot itself and look at
> > > traverse bootenv by all possible dev numbers in sunxi-common.h, but  
> >
> > OK, I will have a look at how the automatic distro boot behaves with
> > this change. My concern was more the interactive user, who is used to
> > have the SD card at device 0 (I certainly am).
> >  
> > > this has one slide effect that we mark mmc2 as devnum 1 for the sake
> > > of fastboot so if we mark fastboot number for specific board properly
> > > (by static or runtime) then explicit aliases wouldn't required.  
> >
> > Ah, good point, thanks for the heads up. I guess this is the actual
> > reason for the alias in our -u-boot.dtsi? Maybe we find a different way
> > to let fastboot find the eMMC? Then we can drop the extra mmc1 alias,
> > get our numbering back, and can cope with the incoming aliases from the
> > mainline DTs as well?  
> 
> Agreed.

So I had a look at fastboot, and posted an RFC[1] to try and fix this
properly. This seems to somewhat work, but there are more assumptions
in (sunxi) U-Boot about the eMMC being device 1 (boot source,
distro_boot scripts, ...). I agree that we should remove this
assumption, but this requires more work and scrutiny to find all those
cases.

For now the change in numbering breaks right, left, and centre: boot.scr
cannot be found anymore, fastboot cannot find the SD card, custom
scripts stop working.

So I would very much like to merge this patch here, to fix the
currently bad user experience and get a clean v2021.07 release.
We can meanwhile look into a proper solution for this problem, to
become more robust against enumeration changes (for instance due to
aliases).

Thanks,
Andre