Re: [PATCH] mm/vmpressure: remove window size TODO

[Lorenzo Stoakes (ARM)]

On Mon, Aug 31, 2026 at 10:16:36AM -0700, Andrew Morton wrote:
> On Mon, 31 Aug 2026 14:42:11 +0100 "Lorenzo Stoakes (ARM)" <ljs@xxxxxxxxxx> wrote:
>
> > There has been a steady stream of patches that have been submitted by
> > newcomers to core mm 'fixing' this TODO, with all but the original having
> > very likely been generated by LLMs.
> >
> > It appears that TODOs to LLMs are like red rags to a bull.
> >
> > In addition, TODOs in code often bitrot and are distracting - those who
> > understand the code know what could be improved in future.
> >
> > Therefore remove the TODO.
> >
> > The work required to actually fix this TODO requires somebody who both has
> > understanding of the code and significant real-world data to back their
> > changes.
> >
> > Such a person doesn't require a TODO prompt to implement this change, so
> > nothing of value is being lost here.
> >
> > ...
> >
> > --- a/mm/vmpressure.c
> > +++ b/mm/vmpressure.c
> > @@ -30,9 +30,6 @@
> >   *
> >   * As the vmscan reclaimer logic works with chunks which are multiple of
> >   * SWAP_CLUSTER_MAX, it makes sense to use it for the window size as well.
> > - *
> > - * TODO: Make the window size depend on machine size, as we do for vmstat
> > - * thresholds. Currently we set it to 512 pages (2MB for 4KB pages).
> >   */
> >  const unsigned long vmpressure_win = SWAP_CLUSTER_MAX * 16;
>
> otoh, we could enhance the TODO to explain these considerations.  This
> might result in the LLM warriors submitting more usable work.
>
> It would be a strange place to socialize these things but as you said,
> red rags.

I don't think it'd help anything, I doubt the LLMs are going to read what
amounts to 'experts only' and sit back, they're not known for their
humility.

I also think, in general, we really should try to avoid adding TODOs - it's
bitrot bait and if it's important enough to do, that should be abundantly
clear to smart developers.

And if people REALLY want to signal that, put it in the commit msg (and
yeah that risks LLM incursion too but we do have to stop somewhere :)

>

--
Cheers, Lorenzo