On Mon, Feb 13, 2023 at 04:28:05PM +0100, Alexandre Ghiti wrote: > The issue lies here: before using pteval in pfn_folio(), we should test it. > The following patch fixes the issue for me: Thanks for debugging it. I'd rather see this written as ... pte_t pteval = *_pte; + unsigned long pfn; + if (pte_none(pteval)) + continue; + pfn = pte_pfn(pteval); + if (is_zero_pfn(pfn)) + continue; + folio = pfn_folio(pfn); + if (folio_test_large(folio)) + continue; release_pte_folio(folio); makes sense? > diff --git a/mm/khugepaged.c b/mm/khugepaged.c > index eb38bd1b1b2f..fef3414b481b 100644 > --- a/mm/khugepaged.c > +++ b/mm/khugepaged.c > @@ -514,10 +514,12 @@ static void release_pte_pages(pte_t *pte, pte_t *_pte, > while (--_pte >= pte) { > pte_t pteval = *_pte; > > - folio = pfn_folio(pte_pfn(pteval)); > - if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval)) && > - !folio_test_large(folio)) > - release_pte_folio(folio); > + if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval))) { > + folio = pfn_folio(pte_pfn(pteval)); > + > + if (!folio_test_large(folio)) > + release_pte_folio(folio); > + } > } > > list_for_each_entry_safe(folio, tmp, compound_pagelist, lru) { > > > @Marek: could you give it a try? > > I can send a separate patch if needed, let me know. > > Thanks, > > Alex > > > > > if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval)) && > > > - !PageCompound(page)) > > > - release_pte_page(page); > > > + !folio_test_large(folio)) > > > + release_pte_folio(folio); > > > } > > > - list_for_each_entry_safe(page, tmp, compound_pagelist, lru) { > > > - list_del(&page->lru); > > > - release_pte_page(page); > > > + list_for_each_entry_safe(folio, tmp, compound_pagelist, lru) { > > > + list_del(&folio->lru); > > > + release_pte_folio(folio); > > > } > > > } > > Best regards >