Peter Xu <peterx@xxxxxxxxxx> writes:
> Ives van Hoorne from codesandbox.io reported an issue regarding possible
> data loss of uffd-wp when applied to memfds on heavily loaded systems. The
> sympton is some read page got data mismatch from the snapshot child VMs.
>
> Here I can also reproduce with a Rust reproducer that was provided by Ives
> that keeps taking snapshot of a 256MB VM, on a 32G system when I initiate
> 80 instances I can trigger the issues in ten minutes.
>
> It turns out that we got some pages write-through even if uffd-wp is
> applied to the pte.
>
> The problem is, when removing migration entries, we didn't really worry
> about write bit as long as we know it's not a write migration entry. That
> may not be true, for some memory types (e.g. writable shmem) mk_pte can
> return a pte with write bit set, then to recover the migration entry to its
> original state we need to explicit wr-protect the pte or it'll has the
> write bit set if it's a read migration entry.
>
> For uffd it can cause write-through. I didn't verify, but I think it'll be
> the same for mprotect()ed pages and after migration we can miss the sigbus
> instead.
>
> The relevant code on uffd was introduced in the anon support, which is
> commit f45ec5ff16a7 ("userfaultfd: wp: support swap and page migration",
> 2020-04-07). However anon shouldn't suffer from this problem because anon
> should already have the write bit cleared always, so that may not be a
> proper Fixes target. To satisfy the need on the backport, I'm attaching
> the Fixes tag to the uffd-wp shmem support. Since no one had issue with
> mprotect, so I assume that's also the kernel version we should start to
> backport for stable, and we shouldn't need to worry before that.
Hi Peter, for the patch feel free to add:
Reviewed-by: Alistair Popple <apopple@xxxxxxxxxx>
I did wonder if this should be backported further for migrate_vma as
well given that a migration failure there might lead a shmem read-only
PTE to become read-write. I couldn't think of an obvious reason why that
would cause an actual problem though.
I think folio_mkclean() will wrprotect the pte for writeback to swap,
but it holds the page lock which prevents migrate_vma installing
migration entries in the first place.
I suppose there is a small window there because migrate_vma will unlock
the page before removing the migration entries. So to be safe we could
consider going back to 8763cb45ab96 ("mm/migrate: new memory migration
helper for use with device memory") but I doubt in practice it's a real
problem.
> Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> Fixes: b1f9e876862d ("mm/uffd: enable write protection for shmem & hugetlbfs")
> Reported-by: Ives van Hoorne <ives@xxxxxxxxxxxxxx>
> Signed-off-by: Peter Xu <peterx@xxxxxxxxxx>
> ---
> mm/migrate.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/mm/migrate.c b/mm/migrate.c
> index dff333593a8a..8b6351c08c78 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -213,8 +213,14 @@ static bool remove_migration_pte(struct folio *folio,
> pte = pte_mkdirty(pte);
> if (is_writable_migration_entry(entry))
> pte = maybe_mkwrite(pte, vma);
> - else if (pte_swp_uffd_wp(*pvmw.pte))
> + else
> + /* NOTE: mk_pte can have write bit set */
> + pte = pte_wrprotect(pte);
> +
> + if (pte_swp_uffd_wp(*pvmw.pte)) {
> + WARN_ON_ONCE(pte_write(pte));
> pte = pte_mkuffd_wp(pte);
> + }
>
> if (folio_test_anon(folio) && !is_readable_migration_entry(entry))
> rmap_flags |= RMAP_EXCLUSIVE;
