Re: [PATCH] mm/hugetlb.c: fix a bug within a BUG(): inconsistent pte comparison

[John Hubbard <jhubbard@xxxxxxxxxx>]

On 6/30/23 03:07, Ryan Roberts wrote:
...
> > diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> > index bce28cca73a1..73fbeb8f979f 100644
> > --- a/mm/hugetlb.c
> > +++ b/mm/hugetlb.c
> > @@ -7246,7 +7246,12 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
> >   				pte = (pte_t *)pmd_alloc(mm, pud, addr);
> >   		}
> >   	}
> > -	BUG_ON(pte && pte_present(ptep_get(pte)) && !pte_huge(ptep_get(pte)));
> > +
> > +	if (pte) {
> > +		pte_t pteval = ptep_get(pte);
>
> Given the PTL is not held here, I think this should technically be
> ptep_get_lockless()?

Oh yes. I'll change it to that and send a v2, thanks!


thanks,
--
John Hubbard
NVIDIA