Re: [PATCH RFC v4 12/12] selftests/mm: add a slow-GUP content and COW test for mTHP

[Rik van Riel <riel@surriel.com>]

On Mon, 2026-07-27 at 16:05 +0200, David Hildenbrand (Arm) wrote:
> On 7/26/26 13:56, Mike Rapoport wrote:
> > 
> > Please (instruct claude to) use ksft_* helpers for printing
> > messages and
> > tracking failed and passed tests.
> 
> I'm afraid we won't have time to keep repeating the following:
> 
> Documentation/process/coding-assistants.rst
> 
> * Reviewing all AI-generated code
> [...]
> * Taking full responsibility for the contribution

I did review the code. I missed this one, largely
because I was focused more on the mm bits.

Not only will I clean it up, I will also make sure
that other people using the available Linux kernel
AI tooling will automatically avoid repeating this
mistake.

I submitted some self test checks to review-prompts:

https://github.com/masoncl/review-prompts/pull/89

And added some commits to kernel-style to create
clearer comments:

https://github.com/rikvanriel/kernel-style

I also tested the patch series against the updated
review-prompts and kernel-style guidelines, and
the issues that were pointed got caught with
the updated tooling.


With the gigablock patch series, I tried to edit
the default LLM output, and create something
vaguely decent. That was a lost cause, which I
do not plan to repeat. 

It was an especially big waste of time with the 
code that I was never planning to keep, but where 
I had not thought of an alternative approach yet.

The current plan of attack is to add LLM rules
every time I get a comment on a patch series
that addresses something generalizable that
I missed.

That way anybody using those tools should have
those mistakes caught, and either raised or
fixed by tooling, before they even start
editing their code.

You don't want to review the same mistakes
over and over again, and I want to help
people avoid sending the same mistakes that
were previously made.

-- 
All Rights Reversed.