Re: [PATCH v2] xarray: clear the new head's parent before freeing the old head

[Lorenzo Stoakes (ARM)]

On Thu, Oct 29, 2026 at 11:50:37AM -0400, Nickolai Zeldovich wrote:
> Hi Lorenzo,
>
> On Thu, Oct 29, 2026 at 11:03 AM Lorenzo Stoakes (ARM) <ljs@xxxxxxxxxx> wrote:
> > It's pretty 101, but please don't send new revisions in-reply-to older
> > revisions or feedback [0].
>
> Understood, will do.  (I was thinking that applied more to
> multi-series patches, which indeed gets unwieldy.)

In general in mm and mm-adjacent we prefer not. I think this might vary by
subsystem, but it gets buried easily in even single patches, so best to default
to not I think.

>
> > Also, if somebody replies to you, reply back. You, not your LLM.
>
> Indeed, I've been replying personally, though in a few cases I sent
> the response in the notes of the v2 patch, in an attempt to send less
> mail to lkml.

OK appreciated!

>
> For context, I use an LLM to write Lean proofs of correctness, which
> seems useful for cases where the patch is non-trivial (e.g., [2]).

Ack, this seems to be a trend. And obviously we're not against LLM usage, as I'm
sure you're aware, this is about rate-limiting that's all.

Also use of lean seems like a nice way of bounding LLMs in a way that makes them
less likely to go awry, I didn't know people were already using that for the
kernel, so that's interesting!

>
> > Also, you are flooding the kernel with patches, 5 series across 5
> > subsystems over 2 hrs:
> >
> > https://lore.kernel.org/all/?q=f%3Anickolai%40csail.mit.edu
> >
> > I already asked you once to stop doing this [1]. You've carried on.
> > [...]
> > I'm sure you're acting in good faith, so can I please ask you to engage
> > with maintainers instead of ignoring us and treating our time as valueless?
>
> Sorry about that.  I'm happy to send bug reports with clear userspace
> reproducers, instead of patches, for any bugs that seem particularly
> significant.  I'll also get in touch with the relevant maintainers to
> see what, if anything, might be useful to them, rather than sending
> individual issues/patches.

In general the bug reports are welcome, just space them out a bit, or if you
have a bunch for a subsystem, contact maintainers ahead of time.

I know you've generally only sent a few-per-subsystem, but in general this is
about managing workload.

Right now the biggest issue with AI submissions is volume-of-submission
vs. review resource. The former can grow exponentially, the latter very much not
:)

>
> Nickolai.
>
> [2] https://lore.kernel.org/lkml/20261027125210.843589-1-nickolai@xxxxxxxxxxxxx/

--
Cheers, Lorenzo