[PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[Suren Baghdasaryan]

When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
obtaining a folio and accessing it even though the entry is swp_entry_t.
Add the missing check and let split_huge_pmd() handle migration entries.

Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Cc: stable@vger.kernel.org
---
Changes since v3 [1]
- Updated the title and changelog, per Peter Xu
- Added Reviewed-by: per Peter Xu

[1] https://lore.kernel.org/all/20250806154015.769024-1-surenb@google.com/

 mm/userfaultfd.c | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)

diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index 5431c9dd7fd7..116481606be8 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -1826,13 +1826,16 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx, unsigned long dst_start,
 			/* Check if we can move the pmd without splitting it. */
 			if (move_splits_huge_pmd(dst_addr, src_addr, src_start + len) ||
 			    !pmd_none(dst_pmdval)) {
-				struct folio *folio = pmd_folio(*src_pmd);
-
-				if (!folio || (!is_huge_zero_folio(folio) &&
-					       !PageAnonExclusive(&folio->page))) {
-					spin_unlock(ptl);
-					err = -EBUSY;
-					break;
+				/* Can be a migration entry */
+				if (pmd_present(*src_pmd)) {
+					struct folio *folio = pmd_folio(*src_pmd);
+
+					if (!folio || (!is_huge_zero_folio(folio) &&
+						       !PageAnonExclusive(&folio->page))) {
+						spin_unlock(ptl);
+						err = -EBUSY;
+						break;
+					}
 				}
 
 				spin_unlock(ptl);

base-commit: 8e7e0c6d09502e44aa7a8fce0821e042a6ec03d1
-- 
2.50.1.565.gc32cd1483b-goog

Re: [PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[David Hildenbrand]

On 07.08.25 00:00, Suren Baghdasaryan wrote:
> When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
> obtaining a folio and accessing it even though the entry is swp_entry_t.
> Add the missing check and let split_huge_pmd() handle migration entries.
> 
> Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
> Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com
> Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/
> Signed-off-by: Suren Baghdasaryan <surenb@google.com>
> Reviewed-by: Peter Xu <peterx@redhat.com>
> Cc: stable@vger.kernel.org
> ---
> Changes since v3 [1]
> - Updated the title and changelog, per Peter Xu
> - Added Reviewed-by: per Peter Xu
> 
> [1] https://lore.kernel.org/all/20250806154015.769024-1-surenb@google.com/
> 
>   mm/userfaultfd.c | 17 ++++++++++-------
>   1 file changed, 10 insertions(+), 7 deletions(-)
> 
> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> index 5431c9dd7fd7..116481606be8 100644
> --- a/mm/userfaultfd.c
> +++ b/mm/userfaultfd.c
> @@ -1826,13 +1826,16 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx, unsigned long dst_start,
>   			/* Check if we can move the pmd without splitting it. */
>   			if (move_splits_huge_pmd(dst_addr, src_addr, src_start + len) ||
>   			    !pmd_none(dst_pmdval)) {
> -				struct folio *folio = pmd_folio(*src_pmd);
> -
> -				if (!folio || (!is_huge_zero_folio(folio) &&
> -					       !PageAnonExclusive(&folio->page))) {
> -					spin_unlock(ptl);
> -					err = -EBUSY;
> -					break;
> +				/* Can be a migration entry */
> +				if (pmd_present(*src_pmd)) {
> +					struct folio *folio = pmd_folio(*src_pmd);
> +
> +					if (!folio


How could you get !folio here? That only makes sense when calling 
vm_normal_folio_pmd(), no?


-- 
Cheers,

David / dhildenb

Re: [PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[Suren Baghdasaryan]

On Thu, Aug 7, 2025 at 3:31 AM David Hildenbrand <david@redhat.com> wrote:
>
> On 07.08.25 00:00, Suren Baghdasaryan wrote:
> > When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
> > obtaining a folio and accessing it even though the entry is swp_entry_t.
> > Add the missing check and let split_huge_pmd() handle migration entries.
> >
> > Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
> > Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com
> > Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/
> > Signed-off-by: Suren Baghdasaryan <surenb@google.com>
> > Reviewed-by: Peter Xu <peterx@redhat.com>
> > Cc: stable@vger.kernel.org
> > ---
> > Changes since v3 [1]
> > - Updated the title and changelog, per Peter Xu
> > - Added Reviewed-by: per Peter Xu
> >
> > [1] https://lore.kernel.org/all/20250806154015.769024-1-surenb@google.com/
> >
> >   mm/userfaultfd.c | 17 ++++++++++-------
> >   1 file changed, 10 insertions(+), 7 deletions(-)
> >
> > diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> > index 5431c9dd7fd7..116481606be8 100644
> > --- a/mm/userfaultfd.c
> > +++ b/mm/userfaultfd.c
> > @@ -1826,13 +1826,16 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx, unsigned long dst_start,
> >                       /* Check if we can move the pmd without splitting it. */
> >                       if (move_splits_huge_pmd(dst_addr, src_addr, src_start + len) ||
> >                           !pmd_none(dst_pmdval)) {
> > -                             struct folio *folio = pmd_folio(*src_pmd);
> > -
> > -                             if (!folio || (!is_huge_zero_folio(folio) &&
> > -                                            !PageAnonExclusive(&folio->page))) {
> > -                                     spin_unlock(ptl);
> > -                                     err = -EBUSY;
> > -                                     break;
> > +                             /* Can be a migration entry */
> > +                             if (pmd_present(*src_pmd)) {
> > +                                     struct folio *folio = pmd_folio(*src_pmd);
> > +
> > +                                     if (!folio
>
>
> How could you get !folio here? That only makes sense when calling
> vm_normal_folio_pmd(), no?

Yes, I think you are right, this check is not needed. I can fold it
into this fix or post a separate cleanup patch. I'm guessing a
separate patch would be better?

>
>
> --
> Cheers,
>
> David / dhildenb
>

Re: [PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[David Hildenbrand]

On 07.08.25 17:27, Suren Baghdasaryan wrote:
> On Thu, Aug 7, 2025 at 3:31 AM David Hildenbrand <david@redhat.com> wrote:
>>
>> On 07.08.25 00:00, Suren Baghdasaryan wrote:
>>> When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
>>> obtaining a folio and accessing it even though the entry is swp_entry_t.
>>> Add the missing check and let split_huge_pmd() handle migration entries.
>>>
>>> Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
>>> Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com
>>> Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/
>>> Signed-off-by: Suren Baghdasaryan <surenb@google.com>
>>> Reviewed-by: Peter Xu <peterx@redhat.com>
>>> Cc: stable@vger.kernel.org
>>> ---
>>> Changes since v3 [1]
>>> - Updated the title and changelog, per Peter Xu
>>> - Added Reviewed-by: per Peter Xu
>>>
>>> [1] https://lore.kernel.org/all/20250806154015.769024-1-surenb@google.com/
>>>
>>>    mm/userfaultfd.c | 17 ++++++++++-------
>>>    1 file changed, 10 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
>>> index 5431c9dd7fd7..116481606be8 100644
>>> --- a/mm/userfaultfd.c
>>> +++ b/mm/userfaultfd.c
>>> @@ -1826,13 +1826,16 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx, unsigned long dst_start,
>>>                        /* Check if we can move the pmd without splitting it. */
>>>                        if (move_splits_huge_pmd(dst_addr, src_addr, src_start + len) ||
>>>                            !pmd_none(dst_pmdval)) {
>>> -                             struct folio *folio = pmd_folio(*src_pmd);
>>> -
>>> -                             if (!folio || (!is_huge_zero_folio(folio) &&
>>> -                                            !PageAnonExclusive(&folio->page))) {
>>> -                                     spin_unlock(ptl);
>>> -                                     err = -EBUSY;
>>> -                                     break;
>>> +                             /* Can be a migration entry */
>>> +                             if (pmd_present(*src_pmd)) {
>>> +                                     struct folio *folio = pmd_folio(*src_pmd);
>>> +
>>> +                                     if (!folio
>>
>>
>> How could you get !folio here? That only makes sense when calling
>> vm_normal_folio_pmd(), no?
> 
> Yes, I think you are right, this check is not needed. I can fold it
> into this fix or post a separate cleanup patch. I'm guessing a
> separate patch would be better?

I think you can just post a fixup inline here and ask Andrew to squash 
it. He will shout if he wants a completely new version :)

-- 
Cheers,

David / dhildenb

Re: [PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[Suren Baghdasaryan]

On Thu, Aug 7, 2025 at 7:42 PM David Hildenbrand <david@redhat.com> wrote:
>
> On 07.08.25 17:27, Suren Baghdasaryan wrote:
> > On Thu, Aug 7, 2025 at 3:31 AM David Hildenbrand <david@redhat.com> wrote:
> >>
> >> On 07.08.25 00:00, Suren Baghdasaryan wrote:
> >>> When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
> >>> obtaining a folio and accessing it even though the entry is swp_entry_t.
> >>> Add the missing check and let split_huge_pmd() handle migration entries.
> >>>
> >>> Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
> >>> Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com
> >>> Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/
> >>> Signed-off-by: Suren Baghdasaryan <surenb@google.com>
> >>> Reviewed-by: Peter Xu <peterx@redhat.com>
> >>> Cc: stable@vger.kernel.org
> >>> ---
> >>> Changes since v3 [1]
> >>> - Updated the title and changelog, per Peter Xu
> >>> - Added Reviewed-by: per Peter Xu
> >>>
> >>> [1] https://lore.kernel.org/all/20250806154015.769024-1-surenb@google.com/
> >>>
> >>>    mm/userfaultfd.c | 17 ++++++++++-------
> >>>    1 file changed, 10 insertions(+), 7 deletions(-)
> >>>
> >>> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> >>> index 5431c9dd7fd7..116481606be8 100644
> >>> --- a/mm/userfaultfd.c
> >>> +++ b/mm/userfaultfd.c
> >>> @@ -1826,13 +1826,16 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx, unsigned long dst_start,
> >>>                        /* Check if we can move the pmd without splitting it. */
> >>>                        if (move_splits_huge_pmd(dst_addr, src_addr, src_start + len) ||
> >>>                            !pmd_none(dst_pmdval)) {
> >>> -                             struct folio *folio = pmd_folio(*src_pmd);
> >>> -
> >>> -                             if (!folio || (!is_huge_zero_folio(folio) &&
> >>> -                                            !PageAnonExclusive(&folio->page))) {
> >>> -                                     spin_unlock(ptl);
> >>> -                                     err = -EBUSY;
> >>> -                                     break;
> >>> +                             /* Can be a migration entry */
> >>> +                             if (pmd_present(*src_pmd)) {
> >>> +                                     struct folio *folio = pmd_folio(*src_pmd);
> >>> +
> >>> +                                     if (!folio
> >>
> >>
> >> How could you get !folio here? That only makes sense when calling
> >> vm_normal_folio_pmd(), no?
> >
> > Yes, I think you are right, this check is not needed. I can fold it
> > into this fix or post a separate cleanup patch. I'm guessing a
> > separate patch would be better?
>
> I think you can just post a fixup inline here and ask Andrew to squash
> it. He will shout if he wants a completely new version :)

I wouldn't do that to him! :)
Let me quickly send an updated version instead.

>
> --
> Cheers,
>
> David / dhildenb
>

Re: [PATCH v4 1/1] userfaultfd: fix a crash in UFFDIO_MOVE when PMD is a migration entry

[Suren Baghdasaryan]

On Thu, Aug 7, 2025 at 7:48 PM Suren Baghdasaryan <surenb@google.com> wrote:
>
> On Thu, Aug 7, 2025 at 7:42 PM David Hildenbrand <david@redhat.com> wrote:
> >
> > On 07.08.25 17:27, Suren Baghdasaryan wrote:
> > > On Thu, Aug 7, 2025 at 3:31 AM David Hildenbrand <david@redhat.com> wrote:
> > >>
> > >> On 07.08.25 00:00, Suren Baghdasaryan wrote:
> > >>> When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
> > >>> obtaining a folio and accessing it even though the entry is swp_entry_t.
> > >>> Add the missing check and let split_huge_pmd() handle migration entries.
> > >>>
> > >>> Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
> > >>> Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com
> > >>> Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/
> > >>> Signed-off-by: Suren Baghdasaryan <surenb@google.com>
> > >>> Reviewed-by: Peter Xu <peterx@redhat.com>
> > >>> Cc: stable@vger.kernel.org
> > >>> ---
> > >>> Changes since v3 [1]
> > >>> - Updated the title and changelog, per Peter Xu
> > >>> - Added Reviewed-by: per Peter Xu
> > >>>
> > >>> [1] https://lore.kernel.org/all/20250806154015.769024-1-surenb@google.com/
> > >>>
> > >>>    mm/userfaultfd.c | 17 ++++++++++-------
> > >>>    1 file changed, 10 insertions(+), 7 deletions(-)
> > >>>
> > >>> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> > >>> index 5431c9dd7fd7..116481606be8 100644
> > >>> --- a/mm/userfaultfd.c
> > >>> +++ b/mm/userfaultfd.c
> > >>> @@ -1826,13 +1826,16 @@ ssize_t move_pages(struct userfaultfd_ctx *ctx, unsigned long dst_start,
> > >>>                        /* Check if we can move the pmd without splitting it. */
> > >>>                        if (move_splits_huge_pmd(dst_addr, src_addr, src_start + len) ||
> > >>>                            !pmd_none(dst_pmdval)) {
> > >>> -                             struct folio *folio = pmd_folio(*src_pmd);
> > >>> -
> > >>> -                             if (!folio || (!is_huge_zero_folio(folio) &&
> > >>> -                                            !PageAnonExclusive(&folio->page))) {
> > >>> -                                     spin_unlock(ptl);
> > >>> -                                     err = -EBUSY;
> > >>> -                                     break;
> > >>> +                             /* Can be a migration entry */
> > >>> +                             if (pmd_present(*src_pmd)) {
> > >>> +                                     struct folio *folio = pmd_folio(*src_pmd);
> > >>> +
> > >>> +                                     if (!folio
> > >>
> > >>
> > >> How could you get !folio here? That only makes sense when calling
> > >> vm_normal_folio_pmd(), no?
> > >
> > > Yes, I think you are right, this check is not needed. I can fold it
> > > into this fix or post a separate cleanup patch. I'm guessing a
> > > separate patch would be better?
> >
> > I think you can just post a fixup inline here and ask Andrew to squash
> > it. He will shout if he wants a completely new version :)
>
> I wouldn't do that to him! :)
> Let me quickly send an updated version instead.

Update posted at
https://lore.kernel.org/all/20250807200418.1963585-1-surenb@google.com/

>
> >
> > --
> > Cheers,
> >
> > David / dhildenb
> >