Bennie/mraa pinmux fix

[benbrenson]

This pull requests covers roughly the following topics:

• Add patches to mraa for supporting pinmuxing via debugfs rather than mmap'ed access to pad configuration registers
• Add patches (linux-iot2050) for supporting the userspace pinmux feature (available since v5.13)
• Add one patch (linux-iot2050) to adapt the iot20250 device tree to support the userspace pinmux feature

I see that there are a lot of changes to persistent kernel patch series which have changed due to my workflow. Changes only apply to nothing but the commit hashes. During kernel development I fetched the cip kernel and applied all persistent patches and then recreated them with my additional patchset. Please don't hesitate to correct me here!

My patchset starts at "0211-pinctrl-pinmux-add-function-selector-to-pinmux-funct.patch".

[jan-kiszka]

Please generate the kernel patch queue with --no-numbered --no-signature.

[benbrenson]

Ok thanks. That reduced the diff a lot. But there are still some problems. This patch here (0048-*) is completely updated. Seems like some changes of this patch made it already into upstream? Or I took the wrong baseline? Or my workflow breaks things here?

What I did:

git clone https://gitlab.com/cip-project/cip-kernel/linux-cip.git
cd linux-cip && git checkout v5.10.162-cip24

Apply all patches found in patches-5.10/:

find patches-5.10/* | while read file; do cat "$file" | git am ; done

All patches applied without any issues.

After that I did my work and regenerated the new patchseries afterwards:

git format-patch --no-numbered --no-signature v5.10.162-cip24..HEAD -o patches/to_isar/

[benbrenson]

Ok I found the issue.

If I append "-C" to the git format-patch command it works. Seems like the copy-detection was turned off on my system.

git format-patch -C --no-numbered --no-signature v5.10.162-cip24..HEAD -o patches/to_isar/ 

This can also permanently enabled by:

git config --global diff.renames copies

Furthermore I had to set the "abbrev" correctly to 12. Otherwise the "index" lines in all patches differentiate as well.

[jan-kiszka]

Ready for testing, I would say. Once we are fine, changes should also be proposed upstream.

For the 5.10 backports, we could think about proposing them for the CIP kernel. OTOH, they are all part of 6.1, right? That will be our next kernel anyway.

[benbrenson]

Ready for testing, I would say. Once we are fine, changes should also be proposed upstream.

For the 5.10 backports, we could think about proposing them for the CIP kernel. OTOH, they are all part of 6.1, right? That will be our next kernel anyway.

Yes. The pinctrl backports have been picked from v5.13. So they can be removed for 6.1.

[jan-kiszka]

@benbrenson please factor out the application of --zero-commit into a separate patch that only does that.

[jan-kiszka]

Looks good otherwise.

[benbrenson]

Sorry for the late reply. I missed your comment and I was sick last week.

I rebased to current master and realigned the patchseries. Meanwhile BaochengSu already added the "--zero-commit" option, so I used that as well.

[jan-kiszka]

Roadmap should likely be now:

1. merge
2. test extensively (system test)
3. propose changes upstream (kernel, mraa)