IB/mlx5: use kvmalloc_array for mlx5_ib_wq

We observed multiple times on our Lustre OSS servers that when
the system memory is fragmented, kmalloc() in create_kernel_qp()
could fail order 4/5 allocations while we still have many free pages.

Switch to kvmalloc_array() to allow the operation to contine.
Signed-off-by: NLi Dongyang <dongyang.li@anu.edu.au>
Acked-by: NLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

drivers/infiniband/hw/mlx5/qp.c

@@ -965,11 +965,16 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
 		goto err_free;
 	}
 
-	qp->sq.wrid = kmalloc(qp->sq.wqe_cnt * sizeof(*qp->sq.wrid), GFP_KERNEL);
-	qp->sq.wr_data = kmalloc(qp->sq.wqe_cnt * sizeof(*qp->sq.wr_data), GFP_KERNEL);
-	qp->rq.wrid = kmalloc(qp->rq.wqe_cnt * sizeof(*qp->rq.wrid), GFP_KERNEL);
-	qp->sq.w_list = kmalloc(qp->sq.wqe_cnt * sizeof(*qp->sq.w_list), GFP_KERNEL);
-	qp->sq.wqe_head = kmalloc(qp->sq.wqe_cnt * sizeof(*qp->sq.wqe_head), GFP_KERNEL);
+	qp->sq.wrid = kvmalloc_array(qp->sq.wqe_cnt,
+				     sizeof(*qp->sq.wrid), GFP_KERNEL);
+	qp->sq.wr_data = kvmalloc_array(qp->sq.wqe_cnt,
+					sizeof(*qp->sq.wr_data), GFP_KERNEL);
+	qp->rq.wrid = kvmalloc_array(qp->rq.wqe_cnt,
+				     sizeof(*qp->rq.wrid), GFP_KERNEL);
+	qp->sq.w_list = kvmalloc_array(qp->sq.wqe_cnt,
+				       sizeof(*qp->sq.w_list), GFP_KERNEL);
+	qp->sq.wqe_head = kvmalloc_array(qp->sq.wqe_cnt,
+					 sizeof(*qp->sq.wqe_head), GFP_KERNEL);
 
 	if (!qp->sq.wrid || !qp->sq.wr_data || !qp->rq.wrid ||
 	    !qp->sq.w_list || !qp->sq.wqe_head) {
@@ -981,11 +986,11 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
 	return 0;
 
 err_wrid:
-	kfree(qp->sq.wqe_head);
-	kfree(qp->sq.w_list);
-	kfree(qp->sq.wrid);
-	kfree(qp->sq.wr_data);
-	kfree(qp->rq.wrid);
+	kvfree(qp->sq.wqe_head);
+	kvfree(qp->sq.w_list);
+	kvfree(qp->sq.wrid);
+	kvfree(qp->sq.wr_data);
+	kvfree(qp->rq.wrid);
 	mlx5_db_free(dev->mdev, &qp->db);
 
 err_free:
@@ -998,11 +1003,11 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
 
 static void destroy_qp_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
 {
-	kfree(qp->sq.wqe_head);
-	kfree(qp->sq.w_list);
-	kfree(qp->sq.wrid);
-	kfree(qp->sq.wr_data);
-	kfree(qp->rq.wrid);
+	kvfree(qp->sq.wqe_head);
+	kvfree(qp->sq.w_list);
+	kvfree(qp->sq.wrid);
+	kvfree(qp->sq.wr_data);
+	kvfree(qp->rq.wrid);
 	mlx5_db_free(dev->mdev, &qp->db);
 	mlx5_buf_free(dev->mdev, &qp->buf);
 }

drivers/infiniband/hw/mlx5/srq.c

@@ -196,7 +196,7 @@ static int create_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq,
 	}
 	mlx5_fill_page_array(&srq->buf, in->pas);
 
-	srq->wrid = kmalloc(srq->msrq.max * sizeof(u64), GFP_KERNEL);
+	srq->wrid = kvmalloc_array(srq->msrq.max, sizeof(u64), GFP_KERNEL);
 	if (!srq->wrid) {
 		err = -ENOMEM;
 		goto err_in;
@@ -230,7 +230,7 @@ static void destroy_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq)
 
 static void destroy_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq)
 {
-	kfree(srq->wrid);
+	kvfree(srq->wrid);
 	mlx5_buf_free(dev->mdev, &srq->buf);
 	mlx5_db_free(dev->mdev, &srq->db);
 }