Mastodawn

Greg K-H Apr 23

After 25+ years of kernel development, I was finally forced to touch `mm/` and it was due to a nommu "issue":
https://lore.kernel.org/lkml/2026042334-acutely-unadorned-e05c@gregkh/

As @axboe said the other day, we aren't expecting a box of chocolates:

https://lore.kernel.org/r/[email protected]

but these past weeks have made me feel like someone owes a few of us kernel developers a bunch of whisky at the very least...

[PATCH] mm/gup: honour FOLL_PIN in NOMMU __get_user_pages_locked() - Greg Kroah-Hartman

Jens Axboe Apr 23

@gregkh I'll take a whisky over a box of chocolates anyway ;-)

Jens Axboe Apr 23

@gregkh Actually I can buy my own, what I'd really like is for us to kill nommu. It has zero reason to exist, and it's largely untested garbage.

Forget the booze, that's what I want.

@axboe @gregkh something @ljs would most likely agree with 🙂 cc @vbabka

@Aissen @gregkh @ljs @vbabka As any sane person should! Honestly, I'm hope this will be the silver lining to the slew of mostly garbage LLM bug reports - yank the old and untested code from the kernel, not just nommu.

@axboe @gregkh @ljs @vbabka yes, I did read https://lwn.net/Articles/1068928/ ; it does seem bittersweet though.

Kernel code removals driven by LLM-created security reports

There are a number of ongoing efforts to remove kernel code, mostly from the networking subsyst [...]

LWN.net

Jens Axboe Apr 23

@Aissen @gregkh @ljs @vbabka Ah had not even seen that. It's not bitter sweet, removing code is a Good Thing, and now there's finally a reason to do it that carries some weight. Before this ordeal, it was always punted with "ah well the maintenance burden isn't THAT big". Kill it with fire, git history is forever and if someone steps up to properly maintain a piece of code, it can be brought back.

We're not running a museum.

Jens Axboe Apr 23

@Aissen @gregkh @ljs @vbabka And let me just add, if you're not able or willing to do any work on this, your opinion on whether or not some code should be removed or stay in the kernel is worth absolutely nothing. Just stay silent, accept the decision of those that ARE carrying this maintenance burden.

@axboe @Aissen @gregkh @vbabka yeah I didn't push that all that much because hch isn't a bullshit artist and has done a lot to deserve respect even if he's fantastically wrong there.

But tbh I should have said that.

git log --author="hch" mm/nommu.c gives me a bunch of cleanup patches that happened to touch the file.

Oh and some work from 2015 :)

Harry (Hyeonggon) Yoo Apr 23

@Aissen @axboe @gregkh @ljs @vbabka

bittersweet because people report bugs on features they don't use rather than fixing them and stepping up to help?

Jens Axboe Apr 23

@hyeyoo @ljs @vbabka @gregkh @Aissen That's not bittersweet, that's just laziness. So much easier to run LLMs on code and report issues they often don't even understand, than it is to write a patch.These people don't care about the code, that's not why they are reporting the issues.

Harry (Hyeonggon) Yoo Apr 23

@axboe @ljs @vbabka @gregkh @Aissen

it's really sad that people do that.
what do these people care then?

Jens Axboe Apr 23

@hyeyoo @ljs @vbabka @gregkh @Aissen "Look I found a bug, how great am I"

@axboe @hyeyoo @vbabka @gregkh @Aissen if they can't explain it in detail themselves, then said reports should be dismissed imo

I mean there's a blurry line when you yourself confirm the report I guess.

But no credit unless you can explain it all, in detail.

Sloppers rarely respond much after you ask this kind of stuff :)

Harry (Hyeonggon) Yoo Apr 23

@axboe @ljs @vbabka @gregkh @Aissen

oh geez, that's not adding much value to the project :/

@hyeyoo @axboe @vbabka @gregkh @Aissen harry say fuck

Harry (Hyeonggon) Yoo Apr 23

@ljs @axboe @vbabka @gregkh @Aissen

funk!

Petr Tesařík Apr 23

@hyeyoo @axboe @ljs @vbabka @gregkh @Aissen But it's adding a dominance signal!

Signal       Standard   Action   Comment
───────────────────────────────────────────────────
SIGDOMINANCE Instinct   Shriek   Social dominance.
                                 Chest-drumming may
                                 be also accepted
                                 in some domains.

*** SHRIEK! ***

Harry (Hyeonggon) Yoo Apr 23

@ptesarik @axboe @ljs @vbabka @gregkh @Aissen

see my dominance signal, a claude subscription!

@hyeyoo @axboe @ljs @vbabka @gregkh I was mostly commenting on the context of network drivers being removed. A modern distro kernel ships those old unmaintained drivers and would allow retro-computing with 3com 3c589 card. Maybe the term isn't well chosen, but Linux got the reputation of having so many drivers, maintained for so long. But every good thing must come to an end.

Jens Axboe Apr 23

@Aissen @hyeyoo @ljs @vbabka @gregkh The retro argument is quite typical. Someone should setup a retro tree and maintain it, those drivers and nommu (and 486 support!) can live there. There's zero reason to keep purely retro support in the kernel imho - if it hasn't been sold in products in decades, it need not be in the mainline tree.

@axboe @hyeyoo @ljs @vbabka @gregkh That seems to be the plan on Jacub's side. https://github.com/linux-netdev/mod-orphan

Note: I'm not arguing against any removal, I'm not crazy enough to take on this type of maintenance. I was just commenting on the feeling.

GitHub - linux-netdev/mod-orphan: Linux networking modules removed from the kernel due to lack of maintenance and high bug count

Linux networking modules removed from the kernel due to lack of maintenance and high bug count - linux-netdev/mod-orphan

GitHub

Jens Axboe Apr 23

@Aissen @hyeyoo @ljs @vbabka @gregkh This is the way. Nothing is free, have the people that want to use it pay the cost. The kernel people have enough to do without maintaining hardware they don't even have.

@Aissen @axboe @hyeyoo @vbabka @gregkh yeah I get it but it holds us back.

These things have VERY REAL impact on development and maintenance.

And when you've broken the nommu build by mistake for the 100th time you get tired of it...

@axboe @Aissen @gregkh @vbabka I already tried this, nommu is untested garbage and a hack and I hate it.

https://lore.kernel.org/linux-mm/77ab7[email protected]/

hch of all people said:

"Nommu is a long standing and reasonable
well maintained part of the kernel, why would anyone want to kill it
for no good reason? I know quite a lot of products shipping it."

https://lore.kernel.org/linux-mm/20241[email protected]/

Also read this - https://lore.kernel.org/linux-mm/639c2[email protected]/

I mean honestly I'm surprised Linus allowed a new nommu arch in, and I'd rather we dump all of them.

It's a big maintenance burden, it's untested, and nobody anywhere has the slightest interest in improving any of that.

@david maybe we should isolate it to some section of maintainers (and out of one of mine ;) and treat it the way it deserves to be treated...

Yeah other than that fine with it 🤣

Re: [RFC PATCH v2 00/13] nommu UML - Lorenzo Stoakes

@axboe @Aissen @gregkh @vbabka @david also I note that it doesn't seem that any of the imagined 'many' people who have interests in nommu lift a finger to maintain or test it.

So honestly on that argument alone it should go...

@ljs pr_warn(“this is not a place of honor\n”); at boot #ifdef