Re: [PATCH v2 07/13] mm/debug_vm_pgtable/set_pte/pmd/pud: Don't use set_*_at to update an existing pte entry

["Aneesh Kumar K.V" <aneesh.kumar@xxxxxxxxxxxxx>]

On 8/20/20 8:02 PM, Christophe Leroy wrote:
> Le 19/08/2020 à 15:01, Aneesh Kumar K.V a écrit :
> > set_pte_at() should not be used to set a pte entry at locations that
> > already holds a valid pte entry. Architectures like ppc64 don't do TLB
> > invalidate in set_pte_at() and hence expect it to be used to set 
> > locations
> > that are not a valid PTE.
> >
> > Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@xxxxxxxxxxxxx>
> > ---
> >   mm/debug_vm_pgtable.c | 35 +++++++++++++++--------------------
> >   1 file changed, 15 insertions(+), 20 deletions(-)
> >
> > diff --git a/mm/debug_vm_pgtable.c b/mm/debug_vm_pgtable.c
> > index 76f4c713e5a3..9c7e2c9cfc76 100644
> > --- a/mm/debug_vm_pgtable.c
> > +++ b/mm/debug_vm_pgtable.c
> > @@ -74,15 +74,18 @@ static void __init pte_advanced_tests(struct 
> > mm_struct *mm,
> >   {
> >       pte_t pte = pfn_pte(pfn, prot);
> > +    /*
> > +     * Architectures optimize set_pte_at by avoiding TLB flush.
> > +     * This requires set_pte_at to be not used to update an
> > +     * existing pte entry. Clear pte before we do set_pte_at
> > +     */
> > +
> >       pr_debug("Validating PTE advanced\n");
> >       pte = pfn_pte(pfn, prot);
> >       set_pte_at(mm, vaddr, ptep, pte);
> >       ptep_set_wrprotect(mm, vaddr, ptep);
> >       pte = ptep_get(ptep);
> >       WARN_ON(pte_write(pte));
> > -
> > -    pte = pfn_pte(pfn, prot);
> > -    set_pte_at(mm, vaddr, ptep, pte);
> >       ptep_get_and_clear(mm, vaddr, ptep);
> >       pte = ptep_get(ptep);
> >       WARN_ON(!pte_none(pte));
> > @@ -96,13 +99,11 @@ static void __init pte_advanced_tests(struct 
> > mm_struct *mm,
> >       ptep_set_access_flags(vma, vaddr, ptep, pte, 1);
> >       pte = ptep_get(ptep);
> >       WARN_ON(!(pte_write(pte) && pte_dirty(pte)));
> > -
> > -    pte = pfn_pte(pfn, prot);
> > -    set_pte_at(mm, vaddr, ptep, pte);
> >       ptep_get_and_clear_full(mm, vaddr, ptep, 1);
> >       pte = ptep_get(ptep);
> >       WARN_ON(!pte_none(pte));
> > +    pte = pfn_pte(pfn, prot);
> >       pte = pte_mkyoung(pte);
> >       set_pte_at(mm, vaddr, ptep, pte);
> >       ptep_test_and_clear_young(vma, vaddr, ptep);
> > @@ -164,9 +165,6 @@ static void __init pmd_advanced_tests(struct 
> > mm_struct *mm,
> >       pmdp_set_wrprotect(mm, vaddr, pmdp);
> >       pmd = READ_ONCE(*pmdp);
> >       WARN_ON(pmd_write(pmd));
> > -
> > -    pmd = pmd_mkhuge(pfn_pmd(pfn, prot));
> > -    set_pmd_at(mm, vaddr, pmdp, pmd);
> >       pmdp_huge_get_and_clear(mm, vaddr, pmdp);
> >       pmd = READ_ONCE(*pmdp);
> >       WARN_ON(!pmd_none(pmd));
> > @@ -180,13 +178,11 @@ static void __init pmd_advanced_tests(struct 
> > mm_struct *mm,
> >       pmdp_set_access_flags(vma, vaddr, pmdp, pmd, 1);
> >       pmd = READ_ONCE(*pmdp);
> >       WARN_ON(!(pmd_write(pmd) && pmd_dirty(pmd)));
> > -
> > -    pmd = pmd_mkhuge(pfn_pmd(pfn, prot));
> > -    set_pmd_at(mm, vaddr, pmdp, pmd);
> >       pmdp_huge_get_and_clear_full(vma, vaddr, pmdp, 1);
> >       pmd = READ_ONCE(*pmdp);
> >       WARN_ON(!pmd_none(pmd));
> > +    pmd = pmd_mkhuge(pfn_pmd(pfn, prot));
> >       pmd = pmd_mkyoung(pmd);
> >       set_pmd_at(mm, vaddr, pmdp, pmd);
> >       pmdp_test_and_clear_young(vma, vaddr, pmdp);
> > @@ -283,18 +279,10 @@ static void __init pud_advanced_tests(struct 
> > mm_struct *mm,
> >       WARN_ON(pud_write(pud));
> >   #ifndef __PAGETABLE_PMD_FOLDED
>
> Same as below, once set_put_at() is gone, I don't think this #ifndef 
> __PAGETABLE_PMD_FOLDED is still need, should be possible to replace by 
> 'if (mm_pmd_folded())'

I would skip that change in this series because I still haven't worked 
out what it means to have FOLDED PMD with 
CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD.


We should probably push that as a cleanup later and somebody who can 
test that config can do that? Currently i can't boot ppc64 with 
DBUG_VM_PGTABLE enabled on ppc64 because it is all buggy w.r.t rules.

-aneesh