[U-Boot] [PATCH 00/52] Support OF_PLATDATA in TPL, enable RK3368 DRAM init and add RK3368-uQ7

[Heiko Stübner]

Hi Łukasz,

Am Montag, 24. Juli 2017, 11:23:05 CEST schrieb Łukasz Majewski:
> On 07/24/2017 10:59 AM, Heiko Stübner wrote:
> > Am Montag, 24. Juli 2017, 16:38:52 CEST schrieb Andy Yan:
> >> Hi Philipp:
> >> 
> >> On 2017年07月21日 18:55, Dr. Philipp Tomsich wrote:
> >>>> On 21 Jul 2017, at 04:33, Andy Yan <andy.yan at rock-chips.com> wrote:
> >>>> 
> >>>> Hi:
> >>>> 
> >>>> On 2017年07月19日 04:35, Philipp Tomsich wrote:
> >>>>> Here's a series that has a lot going on, but I didn't want to split it
> >>>>> to allow everyone to understand the motivation between some of the
> >>>>> changes to the SPL/TPL framework.  The short summary of this is
> >>>>> "enablement for DRAM init on the RK3368 with everything else that's
> >>>>> necessary to do it using OF_PLATDATA in TPL".
> >>>>> 
> >>>>> Enabling the RK3368 created a few more challenges that I'd expected:
> >>>>>   - I needed different stack locations, text addresses and sizes for
> >>>>>   
> >>>>>     TPL and SPL
> >>>>>   
> >>>>>   - the TPL stage was to use OF_PLATDATA and SPL was to use full
> >>>>>   OF_CONTROL
> >>>>>   - with the RK3368 TPL stage, there was a need to extend the
> >>>>>   back-to-bootrom
> >>>>>   
> >>>>>     support for Rockchip to AArch64
> >>>>>   
> >>>>>   - I wanted to start folding some of the Rockchip specific SPL
> >>>>>   support
> >>>>>   
> >>>>>     (i.e. back-to-bootrom) back into the general SPL framework to
> >>>>>     allow
> >>>>>     using it as a general boot method
> >>>> 
> >>>> I have a confusion here: according to the README.TPL,  TPL is Third
> >>>> Program Loader, which loaded by the SPL. So I think it should run after
> >>>> the SPL. But from RK3188 and RK3368 TPL implementation , they all run
> >>>> before SPL.Maybe it is my misunderstanding, hope to get some
> >>>> guidance.:-)>
> >>> 
> >>> I think you are right and most Rockchip boards have been using this the
> >>> wrong way around.
> > 
> > Actually at first I also had the ordering of SPL -> TPL -> uboot for
> > rk3188
> > but after discussions with Simon and Tom Rini, the opinion was that TPL
> > stands for "tiny" program loader and should run before the full SPL.
> 
> Has something changed since:
> 
> http://www.denx.de/wiki/pub/U-Boot/MiniSummitELCE2013/tpl-presentation.pdf
> 
> I've always thought that we do have SPL -> TPL -> u-boot -> Linux
> 
> If there was a change, then:
> 
> 1. The TPL name is a bit misleading (considering the above
> presentation/status)
> 
> 2. IMHO the code running before SPL (to do some _real_ early init)
> should be named more descriptive like PreSPL.

I've gone back and looked up the previous discussion. It was part of a thread 
where we only included Tom later on to clarify spl/tpl usage, so there is no 
list-reference, but I guess I can just quote the part:

----------------- 8< -------------------
Am Donnerstag, 26. Januar 2017, 12:57:40 CEST schrieben Tom Rini:
> On Thu, Jan 26, 2017 at 07:23:25AM -0700, Simon Glass wrote:
> > +Tom for question below.
> > 
> > Hi Heiko,
> > 
> > On 17 December 2016 at 15:51, Heiko Stübner <heiko at sntech.de> wrote:
> > > Hi,
> > > 
> > > Am Samstag, 17. Dezember 2016, 09:57:51 schrieb Simon Glass:
> > >> I cannot see the .S code in that tree. If it is a few instructions
> > >> then it is fine. But if it is a lot of code I question why it cannot
> > >> be written mostly in C?
> > > 
> > > The .S was just my question to Kever how he envisioned this "preloader"
> > > being implemented. Right now all my parts are fully C-based.
> > > 
> > >> Using TPL seems OK to me. But from my understanding TPL runs before
> > >> SPL. So the sequence should be: TPL -> SPL -> U-Boot. This makes some
> > >> sense because SPL is a more full-featured loader, and TPL is just a
> > >> mini-loader to get the SPL running.
> > > 
> > > hmm, I think SPL -> TPL -> U-Boot would be the correct order :-)
> > > Based on the TPL being the "tertiary program loader" where the SPL only
> > > is the "secondary program loader" and based on the slides from 2013 in
> > > [0].> 
> > That seems wrong to me. A secondary load loads the primary loader,
> > which surely means that the tertiary should load the secondary, i.e.
> > 
> > I think it should be:
> >    2 loads 1 (SPL loads U-Boot)
> >    3 loads 2 which loads 1 (TPL loads SPL which loads U-Boot)
> > 
> > The alternative is:
> >    2 loads 3 which loads 1
> > 
> > which seems really odd.
> > 
> > Note that PowerPC appears to use TPL->SPL->U-Boot when I last looked
> > at it. Of course I could be wrong.
> 
> Yes, naming is hard.  TPL (which is maybe better thought of as Tiny not
> Tertiary) loads SPL loads U-Boot.
----------------- 8< -------------------


Heiko