Re: [RFC] Proposal for Adding a Dedicated Debug Kernel Package to Debian(follow-up from DebCamp on DebConf25)

[Yunseong Kim <ysk@kzalloc.com>]

Hi Bastian,

Thank you for your detailed feedback and for discussing this proposal with the
team.

On 8/2/25 6:21 오전, Bastian Blank wrote:
> Hi
> 
> We talked about this in our last team meeting.  We identified some
> questions and also some concerns.
> 
> On Wed, Jul 09, 2025 at 07:43:35PM +0200, Yunseong Kim wrote:
>> I believe having an easily selectable debug-enabled kernel would be a significant
>> benefit for users, particularly during board bring-up or when troubleshooting
>> kernel-related issues. I’d love to hear your thoughts on whether this might be a
>> direction worth exploring.
> 
> Yes, this might be useful for troubleshooting.  I don't yet understand
> how it would be useful for board bring-up, I would assume this usually
> requires code changes.
> 
> This would also mean we can overall disable many debug options in the
> normal kernel.

I'd also like to reiterate one of the key motivations for this proposal. By
moving debugging features to a separate package, we can disable them in the
default kernel. This would lead to a significant throughput increase,
potentially reducing operational costs for the many groups who deploy Debian's
default kernel in their infrastructure. I believe this benefit makes the effort
of overcoming the technical challenges worthwhile.

>> As a potential reference point, Fedora’s approach to debug kernel configuration
>> could be quite helpful. Their debug-config is available in the Rawhide branch:
>> Link: https://src.fedoraproject.org/rpms/kernel/tree/rawhide
> 
> Sadly this does not look like it actually includes source.  I sincerely
> doubt that they modify all those config files by hand this way.  But you
> can deduct some things.
> 
> Mainly they enable
> - many DEBUG options,
> - many DEBUGFS options,
> - some CONFIGFS options, and
> - some handpicked things.
> 
> Fedora installs the debug kernel as it would any other.  But somehow
> they manage to reduce the priority.
> 
> Problems for Debian current are as follows:
> - We have size constraints on our software builders, so right now we
>   just can't build it.
> - We can not make sure it is not used by default.

I will research potential solutions for these and see how they
might be addressed within the Debian Linux Kernel infrastructure.

> Regards,
> Bastian

Thanks again for your input.

Best regards,
Yunseong Kim