Current implementation of move_pages_pte() copies source and destination
PTEs in order to detect concurrent changes to PTEs involved in the move.
However these copies are also used to unmap the PTEs, which will fail if
CONFIG_HIGHPTE is enabled because the copies are allocated on the stack.
Fix this by using the actual PTEs which were kmap()ed.

Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
Reported-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Cc: stable@vger.kernel.org
---
 mm/userfaultfd.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index e0f1e38ac5d8..dda1c9a3662a 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -1290,8 +1290,8 @@ static int move_pages_pte(struct mm_struct *mm, pmd_t *dst_pmd, pmd_t *src_pmd,
 			spin_unlock(src_ptl);
 
 			if (!locked) {
-				pte_unmap(&orig_src_pte);
-				pte_unmap(&orig_dst_pte);
+				pte_unmap(src_pte);
+				pte_unmap(dst_pte);
 				src_pte = dst_pte = NULL;
 				/* now we can block and wait */
 				folio_lock(src_folio);
@@ -1307,8 +1307,8 @@ static int move_pages_pte(struct mm_struct *mm, pmd_t *dst_pmd, pmd_t *src_pmd,
 		/* at this point we have src_folio locked */
 		if (folio_test_large(src_folio)) {
 			/* split_folio() can block */
-			pte_unmap(&orig_src_pte);
-			pte_unmap(&orig_dst_pte);
+			pte_unmap(src_pte);
+			pte_unmap(dst_pte);
 			src_pte = dst_pte = NULL;
 			err = split_folio(src_folio);
 			if (err)
@@ -1333,8 +1333,8 @@ static int move_pages_pte(struct mm_struct *mm, pmd_t *dst_pmd, pmd_t *src_pmd,
 				goto out;
 			}
 			if (!anon_vma_trylock_write(src_anon_vma)) {
-				pte_unmap(&orig_src_pte);
-				pte_unmap(&orig_dst_pte);
+				pte_unmap(src_pte);
+				pte_unmap(dst_pte);
 				src_pte = dst_pte = NULL;
 				/* now we can block and wait */
 				anon_vma_lock_write(src_anon_vma);
@@ -1352,8 +1352,8 @@ static int move_pages_pte(struct mm_struct *mm, pmd_t *dst_pmd, pmd_t *src_pmd,
 		entry = pte_to_swp_entry(orig_src_pte);
 		if (non_swap_entry(entry)) {
 			if (is_migration_entry(entry)) {
-				pte_unmap(&orig_src_pte);
-				pte_unmap(&orig_dst_pte);
+				pte_unmap(src_pte);
+				pte_unmap(dst_pte);
 				src_pte = dst_pte = NULL;
 				migration_entry_wait(mm, src_pmd, src_addr);
 				err = -EAGAIN;
@@ -1396,8 +1396,8 @@ static int move_pages_pte(struct mm_struct *mm, pmd_t *dst_pmd, pmd_t *src_pmd,
 			src_folio = folio;
 			src_folio_pte = orig_src_pte;
 			if (!folio_trylock(src_folio)) {
-				pte_unmap(&orig_src_pte);
-				pte_unmap(&orig_dst_pte);
+				pte_unmap(src_pte);
+				pte_unmap(dst_pte);
 				src_pte = dst_pte = NULL;
 				/* now we can block and wait */
 				folio_lock(src_folio);
-- 
2.48.1.658.g4767266eb4-goog

On Wed, Feb 26, 2025 at 10:55:09AM -0800, Suren Baghdasaryan wrote:
> Current implementation of move_pages_pte() copies source and destination
> PTEs in order to detect concurrent changes to PTEs involved in the move.
> However these copies are also used to unmap the PTEs, which will fail if
> CONFIG_HIGHPTE is enabled because the copies are allocated on the stack.
> Fix this by using the actual PTEs which were kmap()ed.
> 
> Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI")
> Reported-by: Peter Xu <peterx@redhat.com>
> Signed-off-by: Suren Baghdasaryan <surenb@google.com>
> Cc: stable@vger.kernel.org

Reviewed-by: Peter Xu <peterx@redhat.com>

-- 
Peter Xu