Re: [PATCH 4/4] selftests: livepatch: Add information about minimum kernel support

[Miroslav Benes]

On 2026-06-05 17:14:42+02:00, Petr Mladek wrote:
> On Sun 2026-05-24 20:50:33, Marcos Paulo de Souza wrote:
> 
> > The current livepatch selftests are compatible with kernel 4.12, so add
> > a note about it for future developers willing to contribute with new
> > tests.
> > 
> > Signed-off-by: Marcos Paulo de Souza <mpdesouza@xxxxxxxx>
> > ---
> >  tools/testing/selftests/livepatch/README | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/tools/testing/selftests/livepatch/README b/tools/testing/selftests/livepatch/README
> > index d2035dd64a2b..293f4730b927 100644
> > --- a/tools/testing/selftests/livepatch/README
> > +++ b/tools/testing/selftests/livepatch/README
> > @@ -54,3 +54,6 @@ check_result().  The latter function greps the kernel's ring buffer for
> >  those strings for result comparison.  Other utility functions include
> >  general module loading and livepatch loading helpers (waiting for patch
> >  transitions, sysfs entries, etc.)
> > +
> > +All new tests and test modules MUST be compatible with kernel version 4.12 and
> > +later, including current upstream kernels.
> 
> I do not have a good feeling about this. One thing is to accept fixes/workarounds
> for the backward compatibility. But forcing contributors to keep the
> compatibility might be too demanding.

I agree. This patch should be dropped. I think that the first two of the
series make sense on their own since the cleanup is a reasonable thing
to do in general. The third one is in "why not" area. But this one is
wrong. Sashiko summed it up well.

> I might be wrong. But I think that you would need a pretty old
> userspace to run 4.12 kernel. And most people could not get it easily.
> 
> Honestly, I though more about this effort and I am not sure how
> complicated it would be in the long term.
> 
> I know about two pending patchsets which are breaking the backward
> compatibility:
> 
>   + Better integration of callbacks and shadow variables with the states API [1]
>   + Introduce replace set support [2]
> 
> Especially the 1st one is a pain. It is 19th patches and most of them
> are fixing/reworking the selftests. I think that it was the reason
> why nobody reviewed it yet. And it might be even worse if we wanted
> to make it backward compatible.

I would not be too bothered about it. If keeping the backward
compatibility is difficult, it could be put aside. This patch set is not
that complicated and I think that it can be sold like reasonable changes
on their own.