On Mon, Dec 21, 2020 at 10:36:48PM -0500, Jeff Dike wrote:
> virtnet_set_channels can recursively call cpus_read_lock if CONFIG_XPS
> and CONFIG_HOTPLUG are enabled.
>
> The path is:
> virtnet_set_channels - calls get_online_cpus(), which is a trivial
> wrapper around cpus_read_lock()
> netif_set_real_num_tx_queues
> netif_reset_xps_queues_gt
> netif_reset_xps_queues - calls cpus_read_lock()
>
> This call chain and potential deadlock happens when the number of TX
> queues is reduced.
>
> This commit the removes netif_set_real_num_[tr]x_queues calls from
> inside the get/put_online_cpus section, as they don't require that it
> be held.
>
> Signed-off-by: Jeff Dike <jd...@akamai.com>
Acked-by: Michael S. Tsirkin <m...@redhat.com>
> ---
> drivers/net/virtio_net.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 052975ea0af4..e02c7e0f1cf9 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -2093,14 +2093,16 @@ static int virtnet_set_channels(struct net_device
> *dev,
>
> get_online_cpus();
> err = _virtnet_set_queues(vi, queue_pairs);
> - if (!err) {
> - netif_set_real_num_tx_queues(dev, queue_pairs);
> - netif_set_real_num_rx_queues(dev, queue_pairs);
> -
> - virtnet_set_affinity(vi);
> + if (err){
> + put_online_cpus();
> + goto err;
> }
> + virtnet_set_affinity(vi);
> put_online_cpus();
>
> + netif_set_real_num_tx_queues(dev, queue_pairs);
> + netif_set_real_num_rx_queues(dev, queue_pairs);
> + err:
> return err;
> }
>
> --
> 2.17.1
