[PATCH] mm/madvise: Fix VMA_ITERATOR start position

[Liam R. Howlett]

From: Liam Howlett <Liam.Howlett at oracle.com>

The WARN_ON() in vma_iter_store() detected an invalid VMA iterator
state.  Inspecting the code stack from the report shows that the VMA
iterator is never set to the correct start position.

Setting the initialization of the VMA iterator to use the address
'start' fixes this issue.

Reported-by: syzbot+1d9ec4d1a334ae9e1ca6 at syzkaller.appspotmail.com
Signed-off-by: Liam Howlett <Liam.Howlett at oracle.com>
---

Andrew, this can be squashed into the mm-unstable ("madvise: use
split_vma() instead of __split_vma()")


 mm/madvise.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/madvise.c b/mm/madvise.c
index 7db6622f8293..ca672e37b38c 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -142,7 +142,7 @@ static int madvise_update_vma(struct vm_area_struct *vma,
 	struct mm_struct *mm = vma->vm_mm;
 	int error;
 	pgoff_t pgoff;
-	VMA_ITERATOR(vmi, mm, 0);
+	VMA_ITERATOR(vmi, mm, start);
 
 	if (new_flags == vma->vm_flags && anon_vma_name_eq(anon_vma_name(vma), anon_name)) {
 		*prev = vma;
-- 
2.39.0