[U-Boot] Please pull u-boot-sunxi.git/master

[Siarhei Siamashka]

On Fri, 18 Jul 2014 20:14:11 +0100
Ian Campbell <ijc at hellion.org.uk> wrote:

> Hi Albert,
> 
> Welcome back!
> 
> The following changes since commit 23f23f23d509e8e873797884456070c8a47d72b2:
> 
>   socfpga: Relocate arch common functions away from board (2014-07-05 10:14:46 +0200)
> 
> are available in the git repository at:
> 
>   git://git.denx.de/u-boot-sunxi.git master
> 
> for you to fetch changes up to 7c48b015100eeff0e1bbb766394f7beca23afb48:
> 
>   sunxi: use setbits_le32 to enable the DMA clock (2014-07-18 19:42:25 +0100)
> 
> ----------------------------------------------------------------
> Chen-Yu Tsai (1):
>       sunxi: Add support for using MII phy-s with the GMAC nic
> 
> Hans de Goede (12):
>       sunxi: mksunxiboot: Fix loading of files with a size which is not a multiple of 4
>       sunxi: Fix u-boot-spl.lds to refer to .vectors
>       sunxi: Remove mmc DMA support
>       sunxi: Implement reset_cpu
>       sunxi: Add sun4i support
>       sunxi: Add sun5i support

Can we avoid pushing the sun4i and sun5i patches? They cause clean
sunxi dram code rework more difficult.

Just to give an example. Here is what we have now in u-boot for the
sunxi dram code (a very schematic example):

    #ifdef SUN7I
      do_thing_a
      do_thing_b
    #endif
      do_thing_c

Here is what is done by the patches from Hans:

    #ifdef SUN4I
      do_thing_b
      do_thing_a
    #endif
    #if defined(SUN5I) || defined(SUN7I)
      do_thing_a
      do_thing_b
    #endif
      do_thing_c

And here is what we actually want in the end to support all these SoC
variants (with some tweaks and bugfixes for each of these steps):

      do_thing_a
      do_thing_b
      do_thing_c

The reordering of the 'do_thing_a' and 'do_thing_b' operations is
superfluous. And the fugly resulting ifdef hell does not look like a
great improvement, even though the code kinda does the job.

Basically, if these patches are pushed, we have a bigger mess to clean
compared to what we have now. Also if we have a messy support for sun4,
sun5i and sun7i added first, then any future dram code rework needs to
keep the dram functional on all of these SoC variant after every patch.
This is rather inconvenient.

That's why I'm trying to first focus on the sun7i support alone
with the patches:
    http://lists.denx.de/pipermail/u-boot/2014-July/183981.html
And add the support for the other SoC variants only after the initial
sun7i cleanup is done.

Is anyone really in a hurry to get this stuff pushed out of the
u-boot-sunxi tree to upstream right now? Can't we get a 'testing'
branch in the sunxi repository and have everything reviewed first?

-- 
Best regards,
Siarhei Siamashka