On Thu, 10 Mar 2022 18:12:21 +0300, Dan Carpenter <dan.carpenter at
oracle.com> wrote:> Hello Xuan Zhuo,
>
> The patch 26ae35c46f93: "virtio_net: support rx/tx queue reset"
from
> Mar 8, 2022, leads to the following Smatch static checker warning:
Yes, thanks to you, I also found this problem today.
>
> drivers/net/virtio_net.c:1410 virtnet_napi_tx_disable()
> warn: sleeping in atomic context
>
> drivers/net/virtio_net.c
> 1829 static int virtnet_tx_vq_reset(struct virtnet_info *vi,
> 1830 struct send_queue *sq, u32
ring_num)
> 1831 {
> 1832 struct netdev_queue *txq;
> 1833 int err, qindex;
> 1834
> 1835 qindex = sq - vi->sq;
> 1836
> 1837 txq = netdev_get_tx_queue(vi->dev, qindex);
> 1838 __netif_tx_lock_bh(txq);
> ^^^^^^^^^^^^^^^^^^^^^^^
> Disables preempt
>
> 1839
> 1840 /* stop tx queue and napi */
> 1841 netif_stop_subqueue(vi->dev, qindex);
> 1842 virtnet_napi_tx_disable(&sq->napi);
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> napi_disable() is a might_sleep() function.
>
> 1843
> 1844 __netif_tx_unlock_bh(txq);
> 1845
> 1846 /* reset the queue */
> 1847 err = virtio_reset_vq(sq->vq);
> 1848 if (err) {
> 1849 netif_start_subqueue(vi->dev, qindex);
> 1850 goto err;
> 1851 }
>
> regards,
> dan carpenter