Re: [PATCH] mm/khugepaged: reject shmem collapse ranges beyond i_size

[Roman Gushchin <roman@kernel.org>]

"David Hildenbrand (Arm)" <david@kernel.org> writes:

> On 11/13/26 18:20, Roman Gushchin wrote:
>> "David Hildenbrand (Arm)" <david@kernel.org> writes:
>> 
>>> On 11/13/26 15:25, sashiko-bot@kernel.org wrote:
>>>> Thank you for your contribution! Sashiko AI review found no new issues
>>>> in this patch, but noticed 1 related pre-existing issue it does not
>>>> introduce:
>>>> - [Medium] mm: xarray node memory leak in collapse_file() [1]
>>> Ignore this shit from Sashiko.
>>>
>>> Roman, make it stop.
>> 
>> David, will do.
>> 
>> But also it's not how I want to interact on mailing lists tbh.
>
> Roman, just to be clear, I am talking about preexisting issues. We get bombarded
> with preexisting issue reports. Having more duplicate reports posted to the
> mailing list does not help anybody.
>
> It creates more work for everybody. And the least I need right now is more work,
> believe me.

Totally understand.

>> I understand everybody is stressed and overwhelmed by llms etc. Including
>> myself. If sashiko is not appreciated in mm lands, I can turn it off,
>> no problems from me.
>
> Sashiko is great. The embargoed sending of mails to the list is great. I
> appreciate all of that.
>
> Posting preexisting stuff to the mailing list is not great, at least
> not right now.

Yet I've been explicitly asked about it by others. So at least there are
various opinions here.

> We were very happy that preexisting issues went into the private bugtracker. I
> want to delegate stuff to people in my team to screen them and see if it needs
> fixing. Well controlled.
>
> Now either:
>
> (a) it was never really gone, and we were in disbelieve that preexisting issues
>     are no longer posted publicly.
>
> (b) Sashiko maintainer decided to re-enable it without consulting us.

The correct version is b*, it's indeed back, but in a completely
reworked way, as I wrote to Lorenzo yesterday.

Pre-existing reports are now limited to the first non-rfc version of the
patchset, they are limited in numbers per patch (1-2), they are
_deduplicated and posted strictly once per issue_.

Re consulting the problem is that there are several hundreds maintainers
with drastically different views. If we'll wait for everybody's agreement
each time, nothing will ever move. So I have to make some decisions like
this on my own. And, yes, some of them will be wrong or not perfect.
I do fix/adjust things constantly. But some patience would be nice.

Btw now you can delegate access to bugs to people outside of MAINTAINERS.

> I'm sorry if I made you angry, but I hope you can see how I suspect (b) and why
> it would make me angry.

No problems. I understand the problem with the maintainers load right
now.

Thanks