When disconnecting, it's possible that another thread has already made it
into eth_start_xmit before we call netif_stop_queue. This can lead to a
crash as eth_start_xmit tries to use resources that gether_disconnect is
freeing. Use netif_tx_lock/unlock around netif_stop_queue to ensure no
threads are executing during the remainder of gether_disconnect.
Signed-off-by: Jeff Westfahl <[email protected]>
Tested-by: Jaeden Amero <[email protected]>
Signed-off-by: Felipe Balbi <[email protected]>
DBG(dev, "%s\n", __func__);
+ netif_tx_lock(dev->net);
netif_stop_queue(dev->net);
+ netif_tx_unlock(dev->net);
+
netif_carrier_off(dev->net);
/* disable endpoints, forcing (synchronous) completion