Re: [dunfell/master PATCH] ti-sci-fw: k3-image-gen: Revert SPL address change for mainline

Nishanth Menon

On 20:09-20210629, Denys Dmytriyenko wrote:
On Tue, Jun 29, 2021 at 03:36:30PM -0500, Nishanth Menon via wrote:
Commit 02a2433c ("ti-sci-fw: Update image gen to latest") switched to a
newer commit in k3-image-gen that also picked up TISCI firmware release,
However, does include a change where the SPL address changed.

This switch in address has been queued for upstream in a future release
(v5.14 + u-boot 2021.10) and current mainline formal tags do not support
this address change. As a result, we have a broken AM64x mainline

Revert this specific change from k3-image-gen for mainline alone for

NOTE: This uses ARAGO_BRAND=mainline which is NOT accurate for meta-ti
which is distro independent, hence without appropriate changes while
using alternate distros, this change will not take effect.
Right, this flag is Arago Distro specific and it won't do anything for
people using meta-ti directly. Hopefully this can be fixed properly very
soon and this hack removed.
Yes, this is temporary - at least for one u-boot and kernel window :(

One more cleaner approach would be to have a separate "mainline" branch
for k3-image-gen repo, where you have the correct address and mainline
recipe using it - that way it will work for meta-ti by itself...
Another cleaner solution could be to pass SBL_LOADADDR from the recipe
itself to override the one weakly-assigned in the Makefile...
Honestly, I am really hoping this entire k3-image-gen stuff gets
into u-boot as buildman based solution and we should get rid of this
recipe. But, I dont have a clear timeframe for that yet.

I am hoping this is an aberration and should'nt be made as a standard
/ legitimate option by making a valid -mainline recipe or formalize a
framework to provide specific override option- hence the attempt to
engineer it as little as possible.

Let me know if we have a strong opinion on this, and I can try to
respin this.

Meanwhile, I would suggest weakly-definint ARAGO_BRAND in the recipe,
as otherwise the variable will be undefined and may break parsing, i.e.
"undefined behavior":
Yes, this definitely makes sense.


Nishanth Menon
Key (0xDDB5849D1736249D)/Fingerprint: F8A2 8693 54EB 8232 17A3 1A34 DDB5 849D 1736 249D

Join to automatically receive all group messages.