On Mon, 22 May 2023, Qi Zheng wrote:
> On 2023/5/22 12:58, Hugh Dickins wrote:
> > map_pte() use pte_offset_map_nolock(), to make sure of the ptl belonging
> > to pte, even if pmd entry is then changed racily: page_vma_mapped_walk()
> > use that instead of getting pte_lockptr() later, or restart if map_pte()
> > found no page table.
> >
> > Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
> > ---
> > mm/page_vma_mapped.c | 28 ++++++++++++++++++++++------
> > 1 file changed, 22 insertions(+), 6 deletions(-)
> >
> > diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
> > index 947dc7491815..2af734274073 100644
> > --- a/mm/page_vma_mapped.c
> > +++ b/mm/page_vma_mapped.c
> > @@ -156,6 +168,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk
> > *pvmw)
> > struct vm_area_struct *vma = pvmw->vma;
> > struct mm_struct *mm = vma->vm_mm;
> > unsigned long end;
> > + spinlock_t *ptl;
> > pgd_t *pgd;
> > p4d_t *p4d;
> > pud_t *pud;
> > @@ -257,8 +270,11 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk
> > *pvmw)
> > step_forward(pvmw, PMD_SIZE);
> > continue;
> > }
> > - if (!map_pte(pvmw))
> > + if (!map_pte(pvmw, &ptl)) {
> > + if (!pvmw->pte)
> > + goto restart;
>
> Could pvmw->pmd be changed? Otherwise, how about just jumping to the
> retry label below?
>
> @@ -205,6 +205,8 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk
> *pvmw)
> }
>
> pvmw->pmd = pmd_offset(pud, pvmw->address);
> +
> +retry:
> /*
> * Make sure the pmd value isn't cached in a register by the
> * compiler and used as a stale value after we've observed a
You're right, that could be done, and that's where I'd have inserted
the label if there were none already. I just thought the fewer goto
labels the better, so reused the restart already there. If you feel
strongly that it's actively misleading, I can certainly make that
change; but it's too rare an occurrence to be worth optimizing for.
Hugh
