From: Or Gerlitz <ogerlitz@xxxxxxxxxxxx> When dealing with netdev unregister events, we just need to know that this is our currently bounded netdev. There's no need to do any further checks/queries. This patch doesn't change any functionality. Signed-off-by: Or Gerlitz <ogerlitz@xxxxxxxxxxxx> Reviewed-by: Mark Bloch <markb@xxxxxxxxxxxx> Signed-off-by: Saeed Mahameed <saeedm@xxxxxxxxxxxx> Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx> --- It is the same patch as was submitted by Saeed [1], but routed to rdma-next instead of mlx5-next. [1] https://patchwork.kernel.org/patch/10720583/ Thanks --- drivers/infiniband/hw/mlx5/main.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 640fb77dad0c..8470ad2eb5b3 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -172,7 +172,6 @@ static int mlx5_netdev_event(struct notifier_block *this, switch (event) { case NETDEV_REGISTER: - case NETDEV_UNREGISTER: write_lock(&roce->netdev_lock); if (ibdev->rep) { struct mlx5_eswitch *esw = ibdev->mdev->priv.eswitch; @@ -181,15 +180,20 @@ static int mlx5_netdev_event(struct notifier_block *this, rep_ndev = mlx5_ib_get_rep_netdev(esw, ibdev->rep->vport); if (rep_ndev == ndev) - roce->netdev = (event == NETDEV_UNREGISTER) ? - NULL : ndev; + roce->netdev = ndev; } else if (ndev->dev.parent == &mdev->pdev->dev) { - roce->netdev = (event == NETDEV_UNREGISTER) ? - NULL : ndev; + roce->netdev = ndev; } write_unlock(&roce->netdev_lock); break; + case NETDEV_UNREGISTER: + write_lock(&roce->netdev_lock); + if (roce->netdev == ndev) + roce->netdev = NULL; + write_unlock(&roce->netdev_lock); + break; + case NETDEV_CHANGE: case NETDEV_UP: case NETDEV_DOWN: { -- 2.19.1
