[RFC] new board VoCore2

[Mauro Condarelli]

On 12/16/19 8:20 AM, Stefan Roese wrote:
> Hi Mauro,
>
> On 15.12.19 11:20, Mauro Condarelli wrote:
>> I am trying to extend support to a new board "VoCore2" whose specs can
>> be found here: https://vocore.io/v2.html
>
> Nice. Thanks for working on this.
>
>> Port is concerning the "ultimate" board which has all connectors in
>> place for SD, USB and Ethernet.
>> This board comes with a paleolithic (1.3.0) version of u-boot and I'm
>> trying to upgrade to a recent one
>> also because I need to implement "automatic upgrade" (possibly using
>> RAUC, but that is another story).
>>
>> Port is working and able to boot Linuz, but it is not ready for
>> inclusion for several reasons
>> and I would like to have expert advice on how to tackle shortcomings, as
>> advised by stefanro on IRC:
>>> 6:49:14 AM - stefanro: mcon: Please submit the patches directly to the
>>> list, with me (and other MIPS / MT7628/88 experts - Daniel etc) on Cc.
>>> If you feel the patches are not ready for integration, then you can
>>> mark them as "RFC" in the subject.
>>
>> Problematic areas I see (side from possible code style issues) are:
>> 1) I added support for a new SPI NOR Flash and that should, probably, be
>> a separate patch.
>
> Yes, please.
Done.

>> 2) I added support for SPI NOR partitioning and I'm unsure if that
>> should be in the "board definition".
>> 3) SPI NOR partitioning is actually duplicated in ENV var (mtdparts) and
>> in Device Tree; I didn't find how to read it directly from DT.
>
> When you pass the MTD partitioning via kernel cmdline "mtdparts=...",
> which can be generated by U-Boot, then there is no need for the DT
> partitioning.
If I understand You correctly this means to completely remove DT
partitioning
information and purely rely on "mtdparts" information.
I was unaware of this possibility.
I'll cross check.
Thanks.

>> 4) Board also has SD, handled via MTK_SD driver, but I've been unable to
>> make it work.
> Then I suggest to exclude it for now.
I strongly suspect the failure is due to complete inability of VoCore2
to handle
MMC/SD power (SD-VCC is hardwired to +3.3V, no way to hard reset SD card
other than complete power-down) and clock.
I had to patch kernel (5.3.0) mtk_sd.c to disable clock gating
(msdc_gate_clock(),
msdc_ungate_clock() and msdc_drv_probe(); a rather ugly patch I won't try to
send upstream; unfortunately u-boot mtk_sd.c looks very different I think
I'll need some help there...) and to add a "regulator-fixed" to DT, but
I'm unsure
if and how it is supported by u-boot.

I will exclude all this from current patch-set (but it is vital for my
project as I
will need to save Environment to MMC to enable boot-counters without
burning
the SPI NOR).

>> 5) Current binary is rather big and I would like to shrink it a bit, if
>> possible.
>>
>> Any hint/criticism/advice would be VERY welcome.
>
> The most important comment I have is, please post the patch inline, so
> that
> we can easily comment on it. Please also take a look at this page for
> hints
> about patch submission:
>
> http://www.denx.de/wiki/view/U-Boot/Patches
Reading it now.
I will reformat patches using a different branch and condensing them
into logically coherent lumps.
It will take some time.

> Thanks,
> Stefan
>
Regards
Mauro