Re: Known and unfixed active data loss bug in MM + XFS with large folios since Dec 2021 (any kernel from 6.1 upwards)

From: Linus Torvalds
Date: Wed Sep 18 2024 - 23:03:53 EST


On Thu, 19 Sept 2024 at 03:43, Dave Chinner <david@xxxxxxxxxxxxx> wrote:
>
> Should we be asking for 6758c1128ceb to be backported to all
> stable kernels then?

I think we should just do the simple one-liner of adding a
"xas_reset()" to after doing xas_split_alloc() (or do it inside the
xas_split_alloc()).

That said, I do also think it would be really good if the 'xa_lock*()'
family of functions also had something like a

WARN_ON_ONCE(xas->xa_node && !xa_err(xas->xa_node));

which I think would have caught this. Because right now nothing at all
checks "we dropped the xa lock, and held xas state over it".

Linus