[U-Boot] [PATCH v2 5/5] board: toradex: add colibri imx8qxp 2gb wb it v1.0b module support

[Peng Fan]

> Subject: Re: [PATCH v2 5/5] board: toradex: add colibri imx8qxp 2gb wb it
> v1.0b module support
> 
> Hi Peng and Stefano
> 
> On Fri, 2019-04-26 at 10:38 +0200, Stefano Babic wrote:
> > Hi Peng,
> >
> > On 26/04/19 04:10, Peng Fan wrote:
> > > Hi Stefano,
> > >
> > > > Subject: Re: [PATCH v2 5/5] board: toradex: add colibri imx8qxp
> > > > 2gb wb it v1.0b module support
> > > >
> > > > Hi Marcel,
> > > >
> > > > On 25/04/19 14:35, Marcel Ziswiler wrote:
> > > > > Hi Stefano
> > > > >
> > > > > On Thu, 2019-04-25 at 12:48 +0200, Stefano Babic wrote:
> > > > > > Hi Marcel,
> > > > > >
> > > > > > On 09/04/19 17:25, Marcel Ziswiler wrote:
> > > > > > > From: Marcel Ziswiler <marcel.ziswiler at toradex.com>
> > > > > > >
> > > > > > > This commit adds initial support for the Toradex Colibri
> > > > > > > iMX8QXP 2GB WB IT V1.0B module. Unlike the V1.0A early
> > > > > > > access samples exclusively booting from SD card, they are
> > > > > > > now strapped to boot from eFuses which are factory fused to
> > > > > > > properly boot from their on-module eMMC. U- Boot supports
> > > > > > > either booting from the on-module eMMC or
> > > > may
> > > > > > > be used for recovery purpose using the universal update
> > > > > > > utility
> > > > > > > (uuu) aka mfgtools 3.0.
> > > > > > >
> > > > > > > Functionality wise the following is known to be working:
> > > > > > > - eMMC and MMC/SD card
> > > > > > > - Ethernet
> > > > > > > - GPIOs
> > > > > > > - I2C
> > > > > > >
> > > > > > > Unfortunately, there is no USB functionality for the i.MX
> > > > > > > 8QXP as of yet.
> > > > > > >
> > > > > > > Signed-off-by: Marcel Ziswiler <marcel.ziswiler at toradex.com
> > > > > > > >
> > > > > > >
> > > > > > > ---
> > > > > > >
> > > > > >
> > > > > > I merged the series and build locally (fine), but Travis
> > > > > > complains and stops with error:
> > > > > >
> > > > > > +cc1: fatal error: opening output file spl/u-boot-spl.cfgout:
> > > > > > No such
> > > > > > file or directory
> > > > > > +compilation terminated.
> > > > > >
> > > > > > Can you take a look at it ?
> > > > >
> > > > > Sure, looks like Peng's commit caceb739ea07 ("imx: build
> > > > > flash.bin for
> > > > > i.MX8") takes SPL for granted while my patchset currently avoids
> > > > > it.
> > > >
> > > > It looks so, yes.
> > > >
> > > > > BTW: I still don't believe SPL makes much sense on i.MX 8X given
> > > > > all the other proprietary parts involved in booting.
> > > >
> > > > SPL makes more sense if it is possible to detect at runtime the HW
> > > > and change the configuration - for i.MX6, this means RAMS
> > > > detection, which boot device is booting, and so on.
> > > >
> > > > On i.MX8 there is a lot of proprietary parts - we lose the
> > > > flexibility of SPL, and most features are lost (or must be
> > > > provided by proprietary code).
> > > > I agree that
> > > > on this platform SPL makes less sense, and i.MX8 should be built
> > > > independently if CONFIG_SPL is set (this is also for
> > > > i.MX6 / MX5, there are boards without SPL and using the DCD image
> > > > to set up the RAM controller).
> > >
> > > The reason we move to use SPL on i.MX8 is that we would like to
> > > avoid bind ATF/OP-TEE/U-Boot into a flat image with hacked offset in
> > > an image.
> > >
> >
> > It seemed I have missed some point. Thanks for clarification. This
> > makes sense.
> 
> OK, I was also not aware of this.
> 
> However, currently I am just happy the current tooling kinda works.
> Which is we can ship stuff to customers and they may use uuu to recover
> bricked modules. So far nobody is talking about OP-TEE and such advanced
> use cases yet.
> 
> On the other hand enabling SPL currently does not seem to work at all on our
> hardware. Neither booting from eMMC nor recovering using uuu.
> That is really the main reason I decided against it at least for now.

In vendor tree, we use SPL to load i.MX8 container image.
To UUU, 1st need to enable usb gadget functions in SPL, then enable container
for the 2nd image. So with uuu, it not work with upstream U-Boot now.

> 
> > > So the bootflow now is
> > > SPL->ATF->OPTEE->ATF->U-Boot
> > >
> > > Without SPL, when generating flash.bin, we have to hack ATF to copy
> > > OP-TEE image from flash.bin to the runtime location.
> >
> > Nevertheless, I understand that it is not strictly required to enable
> > OPTEE to boot the kernel, and in some applications a secure zone is
> > not required. The thing is not that SPL is used here, but to constrain
> > all other users like Marcel to do the same. With i.MX6, even if I
> > strongly suggested to do this to allow run time detection, I let
> > boards without SPL and with just u-boot.imx (with built-in DCD) to
> > flow into mainline - the board maintainer rules as he knows better
> > where the device is used.
> 
> Thanks!
> 
> > So I will prefer that the build assume to have SPL just if SPL is
> > configured and not in any case, letting boards without SPL (like this
> > colibri-mx8
> 
> Colibri iMX8X that is while on the Apalis family we have the Apalis
> iMX8 (e.g. with the i.MX 8QuadMax or i.MX 8QuadPlus) and the new Apalis
> iMX8X with ECC RAM still in the works. Welcome to the i.MX 8 series with the
> i.MX 8 and i.MX 8X families and careful with NXP's brilliant naming scheme
> (;-p).
> 
> > ) to build.
> 
> Don't worry. I believe I found a fix for the issue at hand and will send a patch
> shortly.

ok.

Regards,
Peng.

> 
> > > > > Plus currently SPL
> > > > > actually breaks the USB serial downloader aka recovery mode
> > > > > using the universal update utility (uuu) aka mfgtools 3.0.
> > >
> > > The usb related function for i.MX8 is not ready now.
> >
> > That is ok - it s WIP, it will be merged when ready. I agree with you,
> > this is *not* a reason to avoid SPL.
> 
> Agreed. However, it is kinda painful requiring different U-Boot configuration
> flavours for regular boot vs. recovery. That said we used to previously do this
> on Apalis/Colibri iMX6 as well so it is definitely no show stopper.
> 
> > > we are almost run out
> > > of ocram with SPL DM, thinking to use OF_PLATDATA now, then move to
> > > usb functions.
> >
> > Understood.
> 
> Yeah, I also played with OF_PLATDATA once before however was not entirely
> happy with the result as of yet. I guess WIP patches welcome (;-p).
> 
> > Best regards,
> > Stefano
> 
> Cheers
> 
> Marcel