[meta-xilinx] Kernel targets

[Bruce Ashfield]

On Wed, Feb 12, 2014 at 5:41 PM, Sipke Vriend <sipke.vriend at xilinx.com> wrote:
> Hi Bruce,
>
> On 13/02/2014 12:27 AM, Bruce Ashfield wrote:
>> On Tue, Feb 11, 2014 at 10:23 PM, Sipke Vriend <sipke.vriend at xilinx.com> wrote:
>>> On 12/02/2014 12:04 PM, Philip Balister wrote:
>>>> I've done some builds using bbappends to modify them. Why do you ask?
>>>>
>>>
>>> Just trying to get some usage statistics.
>>> Currently we only have 3.8 linux yocto targets so we want to see if the target type
>>> is of use or not, for future branches.
>>
>> It might be better to not only consider what we've had in the past, but
>> plan to see what can be done in the future to reduce carrying cost,
>> duplication or or leverage new functionality.
>>
>
> Good point.
>
>> i.e. the 3.10 yocto kernel has LTSI integration (and xilnx's work is
>> included there), it has preempt-rt, aufs, edf and a number of other
>> smaller features. It ends up being updated with -stable, CVE and other
>> releases.
>>
>
> I wonder if the current meta-xilinx patching for linux-yocto targets may well
> Undo some of this integration? Currently linux-xlnx.git merges in Linus' tree
> at intervals and so if linux-yocto LTSI integration is not in that "version" of
> Linus' tree meta-xilinx will patch it "out" in our current patch setup, where
> we effectively "match" what is in linux-xlnx.git for a given "version".

LTSI shows up on a discrete set of kernels, since Greg-kh only selects
one every year as the LTSI kernel. This time around it is the 3.10 kernel,
and I just merged his LTSI content into linux-yocto-3.10, and from watching
the traffic I know there's a fair bit of zynq content in that merge.

If you were to bbappend linux-yocto 3.10, then yes, you'd definitely have
some conflicts.

>
>> So as a non-bleeding edge kernel, it has some advantages and doesn't
>> require xilinx to take the whole carrying cost, while leaving development
>> to focus on upstream and development features.
>>
>
> Reducing carrying cost is always welcome. :-)
>
>> bottom line, is that if a feature mix like I mentioned above ends up
>> being done in 3, 5, 10 different kernel repositories, we all know that it
>> means our development efforts are less focused than they could be.
>>
>
> So, would it would be more useful for meta-xilinx to provide
> linux-yocto targets which are effectively what is in the "linux-yocto tree"
> plus a minimal amount of patches (or none), rather than "matching"
> linux-xlnx.git?

This has been one of the suggestions that I've always made. Start small.
Pick a minimal set of features that showcase the board, or features that
you do want to upstream in the not to distant future .. it's a safe middle
ground, and get them into the tree that lands as the default.  That tree
inherits the common features, CVEs, etc, that I was talking about.

It's the whole "in the box" strategy. If you can get up and running quickly,
you've got attention, and moving to a different kernel baseline, is easier
and some won't even feel the need to move.

Development marches on, and patches flow upstream, so there's less
in any tree but the mainline kernel over time.

If someone wants more support, there's the evil commercial linux vendors
or the xilinx development trees.

Just a few random thoughts to consider, opinions vary and there's no perfect
choice.

Bruce

> With current difference between linux-xlnx.git and linux-yocto-xx.git tree,
> This would mean the linux-yocto target would be somewhat "behind"
> linux-xlnx.git in terms of driver capability etc, but any yocto integration
> would be preserved. For Xilinx an added advantage might be that this
> would assist in determining priorities for the upstreaming efforts for
> the rest of linux-xlnx.git. :-)
>
>> Cheers,
>>
>> Bruce
>>
>>> What's your reason for using linux-yocto over linux-xlnx?
>>> What sort of modifications are you needing to make in your bbappend?
>>>
>>>> Philip
>>>>
>>>> On 02/11/2014 08:55 PM, Sipke Vriend wrote:
>>>>> Hi all,
>>>>>
>>>>> Do we have any active users of the meta-xilinx/recipes-kernel.../linux-yocto_xx targets?
>>>>>
>>>>> Cheers
>>>>> Sipke
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> meta-xilinx mailing list
>>>>> meta-xilinx at yoctoproject.org
>>>>> https://lists.yoctoproject.org/listinfo/meta-xilinx
>>>>>
>>>>>
>>>>
>>>
>>>
>>> _______________________________________________
>>> meta-xilinx mailing list
>>> meta-xilinx at yoctoproject.org
>>> https://lists.yoctoproject.org/listinfo/meta-xilinx
>>
>>
>>
>
>



-- 
"Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end"