Re: [PATCH] linux-custom: support vendor directories via HEADERS_INSTALL_EXTRA

["'Jan Kiszka' via isar-users" <isar-users@googlegroups.com>]

On 12.08.24 13:57, Hombourger, Cedric (DI CTO FDS CES LX) wrote:
> On Mon, 2024-08-12 at 11:41 +0000, Moessbauer, Felix (T CED OES-DE)
> wrote:
>> On Sun, 2024-08-11 at 10:10 +0200, 'Jan Kiszka' via isar-users wrote:
>>> On 10.08.24 23:58, 'Cedric Hombourger' via isar-users wrote:
>>>> Some vendor kernels come with additional directories to be
>>>> included
>>>> in the
>>>> linux-headers package in order to support building of their out-
>>>> of-
>>>> tree
>>>> drivers. HEADERS_INSTALL_EXTRA may be set to a list of
>>>> directories
>>>> relative
>>>> to ${S} in any kernel recipes that includes linux-custom.inc
>>>>
>>>
>>> Does this affect more than the legacy Nvidia Jetpack 5 kernel? I
>>> worked
>>> with a lot of vendor kernels but didn't have the need elsewhere so
>>> far.
>>>
>>> Note that we also didn't merge (or even propose?) the ccache
>>> optimization for that horrible and now (JP6) removed kernel overlay
>>> of
>>
>> I proposed this patch in [1], but due to a lack of downstream
>> consumers
>> (except for JP5) and the questionable issue it is solving, it was
>> never
>> merged.
> 
> Thanks for sharing. I had a chance to look at our JP5 layer earlier
> today (was away for the last two weeks) and I now see that it uses sed
> to insert the "nvidia/include" directory to the list of directories to
> ship with linux-headers.
> 
> No doubt that it works and is simple to maintain.
> 
> My concern is that it introduces a hidden contract between our kernel
> recipe and linux-custom. That said chances are that changes to linux-
> custom will cause a build failure for out-of-tree kernel modules if sed
> did not manage to augment the list of directories to ship. We are
> probably fine with that approach but I generally do not like such
> hidden contracts. The other downside is that other layers may opt to go
> with a different approach and I was hoping that we could get Isar to
> offer a unified way to address this use-case even if the number of
> users may be limited (and will hopefully remain limited as we would not
> want vendor kernels to "pollute" kernel trees with extra folders.
> 

I think Felix was referring to the ccache-related changes. Those are
orthogonal to the header installation.

Again my question: does the QCOM kernel have ccache issues as well?
Should we reconsider Felix' patch?

Jan

-- 
Siemens AG, Technology
Linux Expert Center

-- 
You received this message because you are subscribed to the Google Groups "isar-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isar-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/isar-users/71e2c4c9-17ae-4f3c-934d-2851ad905b0f%40siemens.com.