[U-Boot] u-boot-socfpga repository

[David Hawkins]

Hi Dinh,

>> Up until now I have avoided any SoC development kits as
>> I considered the software support to not have matured
>> enough. I consider "mature" code to be code that I can
>> checkout from mainline, where mainline is U-Boot via the
>> Denx repos, and Linux via the Kernel repos.
>
> For Linux, we have done a better job than u-boot. You should
> be able to have most of what you need from kernel.org for the
> Altera Devkits and Terasic SocKit board. The most important
> piece maybe the FPGA manager, otherwise the SOCFPGA platform
> is just any other A9 board.
>
> The FPGA manager is in-flight:
>
> https://lkml.org/lkml/2014/8/1/517

Thanks, this is valuable and encouraging feedback.

> For U-boot, the upstreaming process has been slow. I admit it, but
> it is very high on our to-do list.

One thing Altera needs to understand is that there are numerous
developers out there that are willing to help. If the upstreaming
process is slow, perhaps that is due to lack of openness? I'm
not saying that is the case here, but its a consideration.
There are plenty of people willing to help, and sometimes all
it takes is asking :)

>> Freescale has done this forever, and I hold their
>> processors and code in high regard.
>>
> I used to work at Freescale's doing the i.MX parts. I hope
> these were the processors you had in the mind?

We've been using the PowerPC products, but the iMX parts are
very nice too, they just didn't happen to have the peripheral
combo I needed.

>> Which ones are supported in mainline U-Boot and Linux?
>> What will it take to make it easier for the end-user
>> like myself?
>
> Echoing earlier...There is good Linux support for the Altera
> Cyclone5 and Arria5 devkit and Terasic SoCkit from kernel.org.

Ok, that is good to know, thank-you.

>> Altera developers, please follow Wolfgang's advice.
>
> Wolfgang's advice is valuable and noted. However, it is in Altera's
> best interest that we have 1 central gathering point for all our
> opensource software support.
>
> I maintain a linux-next git repo at rocketboards for patches that have
> been properly reviewed, and acked-by that are destined for kernel.org.
> The logic should follow that I(Altera) would do the same for u-boot
> patches at rocketboards that are destined for mainline u-boot at denx.

As the maintainer of the U-Boot socfpga repository you would still
have the level of control you want. The rocketboards repo would
be a location that could be used to access a clone of the u-boot
mainline, and as Wolfgang mentioned, your u-boot-socfpga-next
development repo can be anywhere you want it.

Keep in mind that git is not centralized like subversion or CVS,
so having a central git repo is really more of a "convention"
than something required by the architecture. As an end-user of
software, the "brand I trust" is U-Boot, so when I want the
latest source for U-Boot, I go to the source. Rocketboards
does not have any brand recognition for me, so its not a
trusted source.

Keep in mind that Altera's track record with NIOS II and Linux
support will cloud the judgement of many users. I never got
to the point of trying uCLinux or Linux on NIOS II as I
have never seen clean support for that processor architecture.
That situation may have changed now, but the Altera NIOS II
U-Boot and Linux brand was tarnished by poor initial support
and openness.

Please do not take any of these comments as negative, or as
a complaint that you are not doing your job, this is merely
feedback from a third party that just wants to be able to
"plug" an SoC into a system and have working U-Boot and Linux,
so that I can concentrate on my own unique hardware/software
layered upon that solid base.

The open-source community really appreciates Altera taking
the time to listen and benefit from our help.

Cheers,
Dave