Re: [edk2-devel] [edk2-platforms][PATCH 1/1] Platforms/RPi3: Restructure platform in preparation for Pi 4

["Pete Batard" <pete@akeo.ie>]

On 2019.10.21 15:39, Leif Lindholm wrote:
> On Mon, Oct 21, 2019 at 03:28:37PM +0100, Pete Batard wrote:
>> Hi Leif,
>>
>> On 2019.10.21 14:46, Leif Lindholm wrote:
>>> On Mon, Oct 21, 2019 at 03:24:47PM +0200, Ard Biesheuvel wrote:
>>>> On Mon, 21 Oct 2019 at 15:09, Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
>>>>>> If anything, I guess we could consider that the non-osi patch should
>>>>>> come first. Still, whatever we do here, as long as only one of non-osi
>>>>>> and platform is applied, builds are going to be broken, and there is no
>>>>>> way to fix that unless you do consider the set of platforms + non-osi as
>>>>>> a single patch.
>>>>>
>>>>> Agreed, this is a egg/chicken problem.
>>>>
>>>> I dealt with this in the past by just making sure the non-osi and
>>>> platform changes are applied at the same time. So it is good to make
>>>> note of this in the cover letter, but other than that, there is no way
>>>> we can apply interdependent changes to two separate repositories at
>>>> the same time without either breaking bisect for one of them, or
>>>> making a huge effort to add temporary code, defines etc that will be
>>>> removed again right after the changes have landed.
>>>
>>> Agreed. My preference would be to treat edk2-non-osi as the chicken,
>>> and edk2-platforms the egg. I could put the requisite edk2-non-osi
>>> hash into the edk2-platforms commit message before pushing, adding a
>>> line like:
>>>
>>> "This commit requires the edk2-non-osi in use to contain commit <hash>
>>> in order to build."
>>>
>>> If I'm feeling nitpicky, that could replace the comment
>>> "No other changes are being applied at this stage."
>>>
>>> Pete: would you be OK with those two changes?
>>
>> Sounds good. Feel free to go ahead with these changes, thanks.
> 
> In that case:
> Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>
> Pushed as 03f36b8fcfb7.
> 
> Thanks!

Great, thanks!

> Do you know when the Pi4 upstreaming is likely to start?

Well, as opposed to what was the case for the Pi 3 when we started 
upstreaming, we don't have a fully working Pi 4 platform yet (or at 
least, there's a fair amount of work left before we can put a checkmark 
against all the major elements we would like to see checked). So what 
we'll probably be aiming at is figure out some kind of minimum viable 
platform that can be officially submitted, that may not do much, but 
that we can then build upon in a more public and official manner.

Now, since there are quite a few people involved on this one, and we 
need to discuss this internally, I doubt we're going to start publicly 
submitting anything Pi 4 related before one week at least. Figuring out 
what we want to submit, or even if we're truly at a stage where we have 
something proper to submit, is the current next step for us, as we 
wanted to make sure mainline was okay with the structure we are planning 
to go with before moving further (which makes the quick turnover on this 
patchset much appreciated!).

For the record, we have a couple of staging repos at:

https://github.com/samerhaj/edk2-platforms/tree/pi4_staging
https://github.com/samerhaj/edk2-non-osi/tree/pi4_staging

as well as Andrei Warkentin's main development tree (over which the 
repos above are based) at:

https://github.com/andreiw/lampone-edk2-platforms/commits/pi4-hack
https://github.com/andreiw/lampone-edk2-platforms/commits/master

Of course, these are very much WIP still, and not something we can use 
to generate a patchset from.

At this stage, I would say that our biggest issue, apart from various 
drivers not being finalized, is that mainline ARM Trusted Firmware has 
recently integrated support for the Pi 4 in a way that is completely 
different from what we went for (because there was no Trusted Firmware 
to start with, Andrei had to figure out his own, which was based on 
extending the Pi 3's). For starters they went for a BL31 only approach: 
https://github.com/ARM-software/arm-trusted-firmware/commit/f5cb15b0c886afaa41c5d3dad8e859b6a41f76ab. 
This means that, depending on how complex it might be to retrofit 
official ATF, we may submit an initial patchset that relies on our own 
binaries, and leave the switching to official ATF done at a later date...

> As you may have seen from my autoresponder, I'm traveling this week
> and next.

Yes. We'll take that into account.

I believe you might also meet with Samer (El-Haj-Mahmoud), who is 
participating in the Pi 4 porting effort (the first repos above are 
his). So feel free to ask for his views as well.

I guess I'll wish you some pleasant and safe travels then.

Regards,

/Pete