[fedora-arm] ac100 kernel (Was: ARM F-14 Branched report: 20110823 changes)

[Peter Robinson]

On Mon, Sep 5, 2011 at 10:42 AM, Gordan Bobic <gordan at bobich.net> wrote:
>  On Mon, 05 Sep 2011 09:52:58 +0200, Mark Wielaard <mark at klomp.org>
>  wrote:
>
>>> AFAIK we have no out of the box Fedora kernels for any ARM platform
>>> anyway, so "out of the box" doesn't really apply. I am pretty sure
>>> not
>>> enough of the required support has been upstreamed, otherwise we
>>> would
>>> be using the upstream kernel. A lot of AC100 specific fixes are in
>>> marvin24's tree on gitorius, and the base is from ChromeOS which is
>>> trying to be compatible with Tegra2 paz00 boards that AC100 is based
>>> on.
>>> Upstream path is thus somewhat complicated (Marc's (marvin24) AC100
>>> tree, upstream to ChromeOS, which _may_ upstream to mainline but I
>>> suspect like Android it will most likely be periodically forked and
>>> maintained separately). Add to that the nvidia's kernel tree which
>>> is
>>> separate again, and we occassionally include fixes from that (e.g.
>>> Henning (woglinde) recently added in the patches to make the AES
>>> engine
>>> work).
>>>
>>> This isn't really a discussion for the Fedora ARM mailing list.
>>> You'd
>>> probably do better to take the discussion to the AC100 list on
>>> launchpad
>>> and join the #ac100 channel on freenode. I doubt enough of the
>>> required
>>> code will make it upstream any time soon.
>>
>> Yes, I got the impression the ARM kernel situation is a little
>> fragmented :{
>
>  Indeed, and a number of people are getting profoundly unhappy with the
>  state of the ARM support in the kernel.

A number of people were profoundly unhappy, its improving a lot since
2.6.39 as Linaro are actively improving it. Device Tree will help that
but it takes time, I believe its now past its worse and there is now a
team committed to improve it.

>> Just to be clear, for Fedora we would have to make sure
>> the ac100 specific patches from the mavin24, chromeos and nvidia
>> kernel
>> trees move towards the lkml/linus tree? Or can we pick up some of the
>> patches and add them to the fedora spec file first for testing?
>
>  Marc's tree is based on the ChromeOS tree, but includes some AC100
>  specific patches. I don't really know how feeding those upstream would
>  work - in theory it should be a case of feeding them back up to ChromeOS
>  tree, but whether ChromeOS changes filter through to mainline I don't
>  know. Like Android, I suspect it's a Google-maintained fork. I have no
>  idea what nvidia's mainline feeding policy is.
>
>  Or to summarize - it's complicated and I don't know the answer.

I would think if someone was to properly re-align them to mainline and
submit them through the proper process they would likely be accepted.
The issue is more if they are small changes against the main tegra
kernel and what else would the changes affect.

There's nothing to stop someone taking Marc's kernel sources and
building a Fedora kernel from them, and then the appropriate image to
put on the AC100.

>> BTW. Would you happen to have (git) URLs for these trees to make it
>> easier to see the diffs against mainline?
>
>  As I said, it's more complicated than that, but you can have a look
>  here:
>
>  http://gitorious.org/ac100
>  http://nv-tegra.nvidia.com/gitweb/?p=linux-2.6.git;a=summary
>
>  IMO the only sane approach to take at the moment is the pragmatic one
>  of using the kernel that works.

I would build one using the working tree as a start.

Peter