userfaultfd: shmem: UFFDIO_COPY: set the page dirty if VM_WRITE is no…

…t set

commit dcf7fe9 upstream.

Set the page dirty if VM_WRITE is not set because in such case the pte
won't be marked dirty and the page would be reclaimed without writepage
(i.e.  swapout in the shmem case).

This was found by source review.  Most apps (certainly including QEMU)
only use UFFDIO_COPY on PROT_READ|PROT_WRITE mappings or the app can't
modify the memory in the first place.  This is for correctness and it
could help the non cooperative use case to avoid unexpected data loss.

Link: http://lkml.kernel.org/r/[email protected]
Reviewed-by: Hugh Dickins <[email protected]>
Cc: [email protected]
Fixes: 4c27fe4 ("userfaultfd: shmem: add shmem_mcopy_atomic_pte for userfaultfd support")
Reported-by: Hugh Dickins <[email protected]>
Signed-off-by: Andrea Arcangeli <[email protected]>
Cc: "Dr. David Alan Gilbert" <[email protected]>
Cc: Jann Horn <[email protected]>
Cc: Mike Kravetz <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Peter Xu <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

mm/shmem.c

@@ -2305,6 +2305,16 @@ static int shmem_mfill_atomic_pte(struct mm_struct *dst_mm,
 	_dst_pte = mk_pte(page, dst_vma->vm_page_prot);
 	if (dst_vma->vm_flags & VM_WRITE)
 		_dst_pte = pte_mkwrite(pte_mkdirty(_dst_pte));
+	else {
+		/*
+		 * We don't set the pte dirty if the vma has no
+		 * VM_WRITE permission, so mark the page dirty or it
+		 * could be freed from under us. We could do it
+		 * unconditionally before unlock_page(), but doing it
+		 * only if VM_WRITE is not set is faster.
+		 */
+		set_page_dirty(page);
+	}
 
 	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
 
@@ -2338,6 +2348,7 @@ static int shmem_mfill_atomic_pte(struct mm_struct *dst_mm,
 	return ret;
 out_release_uncharge_unlock:
 	pte_unmap_unlock(dst_pte, ptl);
+	ClearPageDirty(page);
 	delete_from_page_cache(page);
 out_release_uncharge:
 	mem_cgroup_cancel_charge(page, memcg, false);