IB/mlx5: Reset access mask when looping inside page fault handler

[ Upstream commit 1abe186ed8a6593069bc122da55fc684383fdc1c ]

If page-fault handler spans multiple MRs then the access mask needs to
be reset before each MR handling or otherwise write access will be
granted to mapped pages instead of read-only.

Cc: <[email protected]> # 3.19
Fixes: 7bdf65d ("IB/mlx5: Handle page faults")
Reported-by: Jerome Glisse <[email protected]>
Signed-off-by: Moni Shoua <[email protected]>
Signed-off-by: Leon Romanovsky <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>

drivers/infiniband/hw/mlx5/odp.c

@@ -497,7 +497,7 @@ void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr)
 static int pagefault_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
 			u64 io_virt, size_t bcnt, u32 *bytes_mapped)
 {
-	u64 access_mask = ODP_READ_ALLOWED_BIT;
+	u64 access_mask;
 	int npages = 0, page_shift, np;
 	u64 start_idx, page_mask;
 	struct ib_umem_odp *odp;
@@ -522,6 +522,7 @@ static int pagefault_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
 	page_shift = mr->umem->page_shift;
 	page_mask = ~(BIT(page_shift) - 1);
 	start_idx = (io_virt - (mr->mmkey.iova & page_mask)) >> page_shift;
+	access_mask = ODP_READ_ALLOWED_BIT;
 
 	if (mr->umem->writable)
 		access_mask |= ODP_WRITE_ALLOWED_BIT;