Hello Lendacky, Thomas,
The patch c5aa9e3b8156: "amd-xgbe: Initial AMD 10GbE platform driver"
from Jun 5, 2014 (linux-next), leads to the following Smatch static
checker warning:
drivers/net/ethernet/amd/xgbe/xgbe-drv.c:1209 xgbe_powerdown()
warn: sleeping in atomic context
drivers/net/ethernet/amd/xgbe/xgbe-drv.c
1181 int xgbe_powerdown(struct net_device *netdev, unsigned int caller)
1182 {
1183 struct xgbe_prv_data *pdata = netdev_priv(netdev);
1184 struct xgbe_hw_if *hw_if = &pdata->hw_if;
1185 unsigned long flags;
1186
1187 DBGPR("-->xgbe_powerdown\n");
1188
1189 if (!netif_running(netdev) ||
1190 (caller == XGMAC_IOCTL_CONTEXT && pdata->power_down)) {
1191 netdev_alert(netdev, "Device is already powered down\n");
1192 DBGPR("<--xgbe_powerdown\n");
1193 return -EINVAL;
1194 }
1195
1196 spin_lock_irqsave(&pdata->lock, flags);
^^^^^^^^^^^^^^^^^
Preempt disabled.
1197
1198 if (caller == XGMAC_DRIVER_CONTEXT)
1199 netif_device_detach(netdev);
1200
1201 netif_tx_stop_all_queues(netdev);
1202
1203 xgbe_stop_timers(pdata);
1204 flush_workqueue(pdata->dev_workqueue);
1205
1206 hw_if->powerdown_tx(pdata);
1207 hw_if->powerdown_rx(pdata);
1208
--> 1209 xgbe_napi_disable(pdata, 0);
This warning is correct, but it only showed up because I messed up my
devel version of Smatch last night to make it assume that we almost
always enter into loops at least one time.
The napi_disable() function cannot be called while holding a spinlock.
You would need to enable DEBUG_ATOMIC_SLEEP to see these warnings at
runtime.
1210
1211 pdata->power_down = 1;
1212
1213 spin_unlock_irqrestore(&pdata->lock, flags);
1214
1215 DBGPR("<--xgbe_powerdown\n");
1216
1217 return 0;
1218 }
regards,
dan carpenter
