On Sun, 21 Jul 2024 16:19:42 -0700 Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Sun, 21 Jul 2024 at 14:54, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > - 875fa64577da ("mm/hugetlb_vmemmap: fix race with speculative PFN
> > walkers") is known to cause a performance regression
> > (https://lore.kernel.org/all/3acefad9-96e5-4681-8014-827d6be71c7a@xxxxxxxxxxxxx/T/#mfa809800a7862fb5bdf834c6f71a3a5113eb83ff).
> > Yu has a fix which I'll send along later via the hotfixes branch.
>
> That commit "875fa64577da" does not exist. I think you are talking
> about commit bd225530a4c7, which is in 6.10 and came in from your last
> hotfixes pull.
>
> I'm assuming 875fa64577da was some commit ID it had before getting
> rebased or something?
>
> I'm also a bit confused about why you mention this, since neither the
> cause _nor_ the fix are in this pull request and you have the wrong
> commit ID.
>
Yes, right, sorry. "mm/hugetlb_vmemmap: fix race with speculative PFN
walkers" was promoted into mm-hotfixes after I make this note.
Anyway. We broke a thing and it's being fixed.
