Re: [PATCH] mm: Update mask post pxd_clear_bad()

[Dev Jain <dev.jain@xxxxxxx>]

On 14/03/25 7:57 pm, Yeo Reum Yun wrote:
> Hi Dev,
>
> > > Since pxd_clear_bad() is an operation changing the state of the page tables,
> > > we should call arch_sync_kernel_mappings() post this.
> > >
> > > Fixes: e80d3909be42 ("mm: track page table modifications in __apply_to_page_range()")
> > > Cc: <stable@xxxxxxxxxxxxxxx>
> > > Signed-off-by: Dev Jain <dev.jain@xxxxxxx>
> > > ---
> > >    mm/memory.c | 4 ++++
> > >    1 file changed, 4 insertions(+)
> > >
> > > diff --git a/mm/memory.c b/mm/memory.c
> > > index 78c7ee62795e..9a4a8c710be0 100644
> > > --- a/mm/memory.c
> > > +++ b/mm/memory.c
> > > @@ -2987,6 +2987,7 @@ static int apply_to_pmd_range(struct mm_struct *mm, pud_t *pud,
> > >                        if (!create)
> > >                                continue;
> > >                        pmd_clear_bad(pmd);
> > > +                     *mask = PGTBL_PMD_MODIFIED;
> >
> > Oh well, I guess these should have been *mask |= PGTBL_PMD_MODIFIED.
> >
> >
> > >                }
> > >                err = apply_to_pte_range(mm, pmd, addr, next,
> > >                                         fn, data, create, mask);
> > > @@ -3023,6 +3024,7 @@ static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
> > >                        if (!create)
> > >                                continue;
> > >                        pud_clear_bad(pud);
> > > +                     *mask = PGTBL_PUD_MODIFIED;
> > >                }
> > >                err = apply_to_pmd_range(mm, pud, addr, next,
> > >                                         fn, data, create, mask);
> > > @@ -3059,6 +3061,7 @@ static int apply_to_p4d_range(struct mm_struct *mm, pgd_t *pgd,
> > >                        if (!create)
> > >                                continue;
> > >                        p4d_clear_bad(p4d);
> > > +                     *mask = PGTBL_P4D_MODIFIED;
> > >                }
> > >                err = apply_to_pud_range(mm, p4d, addr, next,
> > >                                         fn, data, create, mask);
> > > @@ -3095,6 +3098,7 @@ static int __apply_to_page_range(struct mm_struct *mm, unsigned long addr,
> > >                        if (!create)
> > >                                continue;
> > >                        pgd_clear_bad(pgd);
> > +                     mask = PGTBL_PGD_MODIFIED;
> > >                }
> > >                err = apply_to_p4d_range(mm, pgd, addr, next,
> > >                                         fn, data, create, &mask);
>
> I don't think this wouldn't need.
> the pXd_clear_bad() is only called at creation of each level of page table,
> and when it clear, the following, apply_to_pXd_range() function would be set
> the make properly via pXd_alloc() and apply_to_pte_range().

Makes sense. But pxd_clear_bad() gets called in case of !pxd_none(), so 
while creating, why would the page containing the page table not be 
none? I believe it should be cleared already?

> Thanks.