[U-Boot] [PATCH v3 06/13] rockchip: rk3399: Add Orangepi RK3399 support

[Jagan Teki]

On Fri, Apr 26, 2019 at 9:08 PM Paul Kocialkowski
<paul.kocialkowski at bootlin.com> wrote:
>
> Hi,
>
> Le vendredi 26 avril 2019 à 20:48 +0530, Jagan Teki a écrit :
> > On Fri, Apr 26, 2019 at 8:42 PM Paul Kocialkowski
> > <paul.kocialkowski at bootlin.com> wrote:
> > > Hi,
> > >
> > > On Fri, 2019-04-26 at 20:31 +0530, Jagan Teki wrote:
> > > > On Fri, Apr 26, 2019 at 8:24 PM Paul Kocialkowski
> > > > <paul.kocialkowski at bootlin.com> wrote:
> > > > > Hi,
> > > > >
> > > > > On Fri, 2019-04-26 at 20:15 +0530, Jagan Teki wrote:
> > > > > > On Fri, Apr 26, 2019 at 8:04 PM Paul Kocialkowski
> > > > > > <paul.kocialkowski at bootlin.com> wrote:
> > > > > > > Hi,
> > > > > > >
> > > > > > > On Fri, 2019-04-26 at 19:57 +0530, Jagan Teki wrote:
> > > > > > > > On Fri, Apr 26, 2019 at 7:47 PM Paul Kocialkowski
> > > > > > > > <paul.kocialkowski at bootlin.com> wrote:
> > > > > > > > > Hi,
> > > > > > > > >
> > > > > > > > > On Thu, 2019-04-25 at 23:04 +0530, Jagan Teki wrote:
> > > > > > > > > > Add initial support for Orangepi RK3399 board.
> > > > > > > > > >
> > > > > > > > > > Specification
> > > > > > > > > > - Rockchip RK3399
> > > > > > > > > > - 2GB/4GB DDR3
> > > > > > > > > > - 16GB eMMC
> > > > > > > > > > - SD card slot
> > > > > > > > >
> > > > > > > > > Looks like you're missing u-boot,dm-pre-reloc to have it working, which
> > > > > > > > > will need to be introduced when moving to rk3399-u-boot.dtsi.
> > > > > > > >
> > > > > > > > Look like you are confused or doesn't check the patch. This patch have
> > > > > > > > rk3399-orangepi-u-boot.dtsi which included rk3399-u-boot.dtsi that has
> > > > > > > > u-boot,dm-pre-reloc for sdmmc.
> > > > > > >
> > > > > > > Well no, in your v3.1 patch, you no longer have u-boot,dm-pre-reloc in
> > > > > > > rk3399-u-boot.dtsi, so you need to add it in the device dts and remove
> > > > > > > it in your second series when you add it back to rk3399-u-boot.dtsi.
> > > > > >
> > > > > > Which u-boot,dm-pre-reloc are you taking about?
> > > > > >
> > > > > > v3.1 has u-boot,dm-pre-reloc for sdmmc, please check it again [1]
> > > > > > which were used by subsequent boards on the same series.
> > > > > >
> > > > > > The diff between v3 vs v3.1 is like v3 removed u-boot,dm-pre-reloc on
> > > > > > existing board dts files thought that it will include
> > > > > > rk3399-u-boot.dtsi will automatically, but not ie fixed in v3.1
> > > > > >
> > > > > > > It's not okay to submit the board with broken MMC support and fix it in
> > > > > > > a subsequent series.
> > > > > >
> > > > > > Again, nothing broken. existing boards or dts(i) files are untouched.
> > > > > > Added only initial rk3399-u-boot.dtsi with sdmmc u-boot,dm-pre-reloc
> > > > > > node to make use of new boards.  and the same reused by next series
> > > > > > so-that I can add binman global to all rk3399 boards.
> > > > >
> > > > > Okay I think I'm getting there. So the Orangepi uses the new scheme
> > > > > (including rk3399-u-boot.dtsi which has u-boot,dm-pre-reloc) while all
> > > > > other boards are still using the previous scheme after the series.
> > > > >
> > > > > This is very confusing and I really think you should keep u-boot,dm-
> > > > > pre-reloc in the orange pi dts file and then make the transition with
> > > > > all the other boards. You're mixing multiple logical steps which makes
> > > > > it really hard to understand what's going on.
> > > > >
> > > > > More to that, introducing the rk3399-u-boot.dtsi is a logical step that
> > > > > should be grouped with your second series, not the first one. In your
> > > > > first series, boards have u-boot,dm-pre-reloc in their per-device dtsi.
> > > > >
> > > > > Could you respin the two series to group changes by logic changes
> > > > > instead of the current state of interleaved changes?
> > > >
> > > > Okay, to make it clear I can send v4 with v3.1 included and updated
> > > > commit log. but I the new series about binman remains same since the
> > > > changes on first patch on the series are relevant.
> > >
> > > I think it makes no sense to introduce rk3399-u-boot.dts as part of the
> > > series currently in v3. You need to remove the patch (currently v3.1)
> > > from this series and group it with the new series (binman-related).
> > >
> > > Otherwise, you're mixing two unrelated logical changes across two
> > > series.
> >
> > rk3399-u-boot.dtsi is require for sdboot on that boards, whole idea is
> > to untouch Linux dts files and get introduced initial
> > rk3399-u-boot.dtsi which doesn't effect any board dts files but use
> > new boards.
>
> Yes I think I properly understand that now. The problem is that
> currently, rk3399 boards do not use rk3399-u-boot.dtsi. In the first
> series, you are introducing the file and including it in the newly-
> introduced devices, but not in the previous ones. This is inconsistent.
>
> Then, in the second series, you are moving existing boards to using
> rk3399-u-boot.dtsi, except for the boards you introduced in the first
> series. This is inconsistent.
>
> Do you see the problem? I feel like you are missing it and only
> focusing on the fact that the series works in practice, which is not
> the issue.

Not again.

First series, I get introduced rk3399-u-boot.dtsi, and only the new
boards are using this file and next series rest of the boards are
using which indeed a valid step. what is inconsistent here, I don't
understand.

>
> > rk3399-u-boot.dtsi is required for binman for adding binman changes,
> > and indeed binman series has some issues which can take time to merge
> > and other one can do it before.
>
> Yes, it can be done before in the first series (so things will not
> break), but it should not be done that way for the reason I'm talking
> about. Again, the way changes are split and organized is as import as
> the technical changes. It's not sufficient that the series works, it
> also has to make sense by grouping logical changes.

Don't agree your point, let me send v4 will see what Philipp and other
can think off.

thanks for your comments,
Jagan.