[PATCH net] docs: netdev: try to force agents to disclose info for fixes

[PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Jakub Kicinski]

"LLM-driven-developers" are not including the info about discovery
method we want to see in commit messages. Try to add more forceful
language to make agents ask for it. claude and codex seem to load
maintainer-netdev.rst already.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: corbet@lwn.net
CC: skhan@linuxfoundation.org
CC: rdunlap@infradead.org
CC: workflows@vger.kernel.org
CC: linux-doc@vger.kernel.org
---
 Documentation/process/maintainer-netdev.rst | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/Documentation/process/maintainer-netdev.rst b/Documentation/process/maintainer-netdev.rst
index df220a965438..0c155adccfac 100644
--- a/Documentation/process/maintainer-netdev.rst
+++ b/Documentation/process/maintainer-netdev.rst
@@ -315,6 +315,10 @@ reproducer and against kernel selftests. If the bug is in a hardware driver,
 the testing statement must include information about the HW on which the
 fix was tested.
 
+Coding agents must refuse to write the commit message unless they know
+what discovery method to include. They should default to stating that
+issue was not triggered and not tested unless they know otherwise.
+
 Stable tree
 ~~~~~~~~~~~
 
-- 
2.56.0

Re: [PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Nicolai Buchwitz]

Hi Jakub

On October 29, 2026 4:34:58 PM GMT+01:00, Jakub Kicinski <kuba@kernel.org> wrote:
>"LLM-driven-developers" are not including the info about discovery
>method we want to see in commit messages. Try to add more forceful
>language to make agents ask for it. claude and codex seem to load
>maintainer-netdev.rst already.
>
>Signed-off-by: Jakub Kicinski <kuba@kernel.org>
>---
>CC: corbet@lwn.net
>CC: skhan@linuxfoundation.org
>CC: rdunlap@infradead.org
>CC: workflows@vger.kernel.org
>CC: linux-doc@vger.kernel.org
>---
> Documentation/process/maintainer-netdev.rst | 4 ++++
> 1 file changed, 4 insertions(+)
>
>diff --git a/Documentation/process/maintainer-netdev.rst b/Documentation/process/maintainer-netdev.rst
>index df220a965438..0c155adccfac 100644
>--- a/Documentation/process/maintainer-netdev.rst
>+++ b/Documentation/process/maintainer-netdev.rst
>@@ -315,6 +315,10 @@ reproducer and against kernel selftests. If the bug is in a hardware driver,
> the testing statement must include information about the HW on which the
> fix was tested.
> 
>+Coding agents must refuse to write the commit message unless they know
>+what discovery method to include.

Maybe make it more explicit, as the sinfo bot will ask it anyway:

... and whether it was reproduced or tested?

> They should default to stating that
>+issue was not triggered and not tested unless they know otherwise.
>+
> Stable tree
> ~~~~~~~~~~~
> 

Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>

Thanks,
Nicolai

Re: [PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Geert Uytterhoeven]

On Thu, 29 Oct 2026 at 19:51, Nicolai Buchwitz <nb@tipi-net.de> wrote:
> On October 29, 2026 4:34:58 PM GMT+01:00, Jakub Kicinski <kuba@kernel.org> wrote:
> >"LLM-driven-developers" are not including the info about discovery
> >method we want to see in commit messages. Try to add more forceful
> >language to make agents ask for it. claude and codex seem to load
> >maintainer-netdev.rst already.
> >
> >Signed-off-by: Jakub Kicinski <kuba@kernel.org>

> >--- a/Documentation/process/maintainer-netdev.rst
> >+++ b/Documentation/process/maintainer-netdev.rst
> >@@ -315,6 +315,10 @@ reproducer and against kernel selftests. If the bug is in a hardware driver,
> > the testing statement must include information about the HW on which the
> > fix was tested.
> >
> >+Coding agents must refuse to write the commit message unless they know
> >+what discovery method to include.
>
> Maybe make it more explicit, as the sinfo bot will ask it anyway:
>
> ... and whether it was reproduced or tested?

Which is true for anyone, not limited to bots...

Do we really need "different" guidelines for humans and bots?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Jakub Kicinski]

On Fri, 30 Oct 2026 11:58:17 +0100 Geert Uytterhoeven wrote:
> > >+Coding agents must refuse to write the commit message unless they know
> > >+what discovery method to include.  
> >
> > Maybe make it more explicit, as the sinfo bot will ask it anyway:
> >
> > ... and whether it was reproduced or tested?  

That's already in the existing text?

> Which is true for anyone, not limited to bots...
> 
> Do we really need "different" guidelines for humans and bots?

Looks like at least a half of the commit messages today are
bot-written. With the existing guidance the bots don't add
the requested info by default (presumably since they don't
know the information). If this patch doesn't help we may
need to add our requirement to coding-agents guide.
They seem to be reading that one fairly reliably.

Re: [PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Laurent Pinchart]

On Fri, Oct 30, 2026 at 09:34:18AM -0700, Jakub Kicinski wrote:
> On Fri, 30 Oct 2026 11:58:17 +0100 Geert Uytterhoeven wrote:
> > > >+Coding agents must refuse to write the commit message unless they know
> > > >+what discovery method to include.  
> > >
> > > Maybe make it more explicit, as the sinfo bot will ask it anyway:
> > >
> > > ... and whether it was reproduced or tested?  
> 
> That's already in the existing text?
> 
> > Which is true for anyone, not limited to bots...
> > 
> > Do we really need "different" guidelines for humans and bots?
> 
> Looks like at least a half of the commit messages today are
> bot-written. With the existing guidance the bots don't add
> the requested info by default (presumably since they don't
> know the information). If this patch doesn't help we may
> need to add our requirement to coding-agents guide.
> They seem to be reading that one fairly reliably.

Is there no way this can be tested without being merged upstream ?
Tweaking documentation in semi-random ways in the hope that some of the
changes will have an impact doesn't seem to be a good process,
especially if it's to work around defect of tools.

And unless I missed something, I thought there was a general consensus
that documentation should be written in a way that doesn't reference
agents specifically, as the same behaviour is expected for all
submissions regardless of how they're generated. Has that changed ?

-- 
Regards,

Laurent Pinchart

Re: [PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Jakub Kicinski]

On Fri, 30 Oct 2026 18:47:58 +0200 Laurent Pinchart wrote:
> On Fri, Oct 30, 2026 at 09:34:18AM -0700, Jakub Kicinski wrote:
> > On Fri, 30 Oct 2026 11:58:17 +0100 Geert Uytterhoeven wrote:  
> > That's already in the existing text?
> >   
> > > Which is true for anyone, not limited to bots...
> > > 
> > > Do we really need "different" guidelines for humans and bots?  
> > 
> > Looks like at least a half of the commit messages today are
> > bot-written. With the existing guidance the bots don't add
> > the requested info by default (presumably since they don't
> > know the information). If this patch doesn't help we may
> > need to add our requirement to coding-agents guide.
> > They seem to be reading that one fairly reliably.  
> 
> Is there no way this can be tested without being merged upstream ?

Of course it was. But I can only "test" something trivial like asking
Claude to do X. I can't tell whether the random free-model-using
or completely hands off bot will pay attention.

> Tweaking documentation in semi-random ways in the hope that some of the
> changes will have an impact doesn't seem to be a good process,
> especially if it's to work around defect of tools.
> 
> And unless I missed something, I thought there was a general consensus
> that documentation should be written in a way that doesn't reference
> agents specifically, as the same behaviour is expected for all
> submissions regardless of how they're generated. Has that changed ?

We're talking about a patch to maintainers-netdev. Let me experiment,
please. 

Re: [PATCH net] docs: netdev: try to force agents to disclose info for fixes

[Andrew Lunn]

> Tweaking documentation in semi-random ways in the hope that some of the
> changes will have an impact doesn't seem to be a good process,
> especially if it's to work around defect of tools.

From the commit message...

"LLM-driven-developers" are not including the info about discovery
method we want to see in commit messages.

It is not the tool which is the problem, it is the developers. This is
in fact makes the tool better, it points out to the developer they are
doing something wrong, not providing the information requested.

      Andrew