Re: [PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[Ralph Campbell]

On 8/26/19 8:11 AM, Vlastimil Babka wrote:
> On 7/20/19 1:32 AM, Ralph Campbell wrote:
> > When CONFIG_MIGRATE_VMA_HELPER is enabled, migrate_vma() calls
> > migrate_vma_collect() which initializes a struct mm_walk but
> > didn't initialize mm_walk.pud_entry. (Found by code inspection)
> > Use a C structure initialization to make sure it is set to NULL.
> >
> > Fixes: 8763cb45ab967 ("mm/migrate: new memory migration helper for use with
> > device memory")
> > Cc: stable@xxxxxxxxxxxxxxx
> > Signed-off-by: Ralph Campbell <rcampbell@xxxxxxxxxx>
> > Cc: "JÃrÃme Glisse" <jglisse@xxxxxxxxxx>
> > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
>
> So this bug can manifest by some garbage address on stack being called, right? I
> wonder, how comes it didn't actually happen yet?

Right.
Probably because HMM isn't widely being used in production yet.

> > ---
> >   mm/migrate.c | 17 +++++++----------
> >   1 file changed, 7 insertions(+), 10 deletions(-)
> >
> > diff --git a/mm/migrate.c b/mm/migrate.c
> > index 515718392b24..a42858d8e00b 100644
> > --- a/mm/migrate.c
> > +++ b/mm/migrate.c
> > @@ -2340,16 +2340,13 @@ static int migrate_vma_collect_pmd(pmd_t *pmdp,
> >   static void migrate_vma_collect(struct migrate_vma *migrate)
> >   {
> >   	struct mmu_notifier_range range;
> > -	struct mm_walk mm_walk;
> > -
> > -	mm_walk.pmd_entry = migrate_vma_collect_pmd;
> > -	mm_walk.pte_entry = NULL;
> > -	mm_walk.pte_hole = migrate_vma_collect_hole;
> > -	mm_walk.hugetlb_entry = NULL;
> > -	mm_walk.test_walk = NULL;
> > -	mm_walk.vma = migrate->vma;
> > -	mm_walk.mm = migrate->vma->vm_mm;
> > -	mm_walk.private = migrate;
> > +	struct mm_walk mm_walk = {
> > +		.pmd_entry = migrate_vma_collect_pmd,
> > +		.pte_hole = migrate_vma_collect_hole,
> > +		.vma = migrate->vma,
> > +		.mm = migrate->vma->vm_mm,
> > +		.private = migrate,
> > +	};
> >   
> >   	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm_walk.mm,
> >   				migrate->start,